fix: replace orchestrator-develop branch references with main

The orchestrator-develop branch no longer exists. Update all fallback
clone commands and test fixtures to use main instead.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

action.yml

@@ -182,8 +182,8 @@ inputs:
     required: false
     default: ''
     description:
-      '[Orchestrator] Run a custom job instead of the standard build automation for orchestrator (in yaml format with the
-      keys image, secrets (name, value object array), command line string)'
+      '[Orchestrator] Run a custom job instead of the standard build automation for orchestrator (in yaml format with
+      the keys image, secrets (name, value object array), command line string)'
   awsStackName:
     default: 'game-ci'
     required: false
@@ -194,42 +194,6 @@ inputs:
     description:
       '[Orchestrator] Either local, k8s or aws can be used to run builds on a remote cluster. Additional parameters must
       be configured.'
-  fallbackProviderStrategy:
-    default: ''
-    required: false
-    description:
-      '[Orchestrator] Fallback provider when the primary is unavailable. Used with runnerCheckEnabled for automatic
-      failover, or as a catch-all if the primary provider fails to initialize.'
-  runnerCheckEnabled:
-    default: 'false'
-    required: false
-    description:
-      '[Orchestrator] Check GitHub Actions runner availability before starting a build. When no suitable runners are
-      available and fallbackProviderStrategy is set, automatically routes to the fallback provider.'
-  runnerCheckLabels:
-    default: ''
-    required: false
-    description:
-      '[Orchestrator] Comma-separated runner labels to filter when checking availability (e.g. self-hosted,linux).
-      When empty, checks all runners in the repository.'
-  runnerCheckMinAvailable:
-    default: '1'
-    required: false
-    description:
-      '[Orchestrator] Minimum number of idle runners required for the primary provider. If fewer are available,
-      routes to fallbackProviderStrategy.'
-  retryOnFallback:
-    default: 'false'
-    required: false
-    description:
-      '[Orchestrator] When true and fallbackProviderStrategy is set, automatically retry the build on the fallback
-      provider if the primary provider fails. Useful for long builds where transient cloud failures are common.'
-  providerInitTimeout:
-    default: '0'
-    required: false
-    description:
-      '[Orchestrator] Maximum seconds to wait for the primary provider to initialize (setupWorkflow). If exceeded
-      and fallbackProviderStrategy is set, switches to the fallback. Set to 0 to disable (default).'
   resourceTracking:
     default: 'false'
     required: false
@@ -316,6 +280,35 @@ inputs:
       '[Orchestrator] Specifies the repo for the unity builder. Useful if you forked the repo for testing, features, or
       fixes.'
 
+  hotRunnerEnabled:
+    description: '[HotRunner] Use persistent hot runner for builds (requires pre-registered runners)'
+    required: false
+    default: 'false'
+  hotRunnerTransport:
+    description: '[HotRunner] Transport protocol for hot runner communication: websocket, grpc, named-pipe'
+    required: false
+    default: 'websocket'
+  hotRunnerHost:
+    description: '[HotRunner] Hot runner host address'
+    required: false
+    default: 'localhost'
+  hotRunnerPort:
+    description: '[HotRunner] Hot runner port number'
+    required: false
+    default: '9090'
+  hotRunnerHealthInterval:
+    description: '[HotRunner] Health check interval in seconds'
+    required: false
+    default: '30'
+  hotRunnerMaxIdle:
+    description: '[HotRunner] Maximum idle time in seconds before recycling runner'
+    required: false
+    default: '3600'
+  hotRunnerFallbackToCold:
+    description: '[HotRunner] Fall back to cold build if no hot runner available'
+    required: false
+    default: 'true'
+
 outputs:
   volume:
     description: 'The Persistent Volume (PV) where the build artifacts have been stored by Kubernetes'

src/index.ts

@@ -3,6 +3,8 @@ import { Action, BuildParameters, Cache, Orchestrator, Docker, ImageTag, Output
 import { Cli } from './model/cli/cli';
 import MacBuilder from './model/mac-builder';
 import PlatformSetup from './model/platform-setup';
+import { HotRunnerService } from './model/orchestrator/services/hot-runner';
+import { HotRunnerConfig } from './model/orchestrator/services/hot-runner/hot-runner-types';
 
 async function runMain() {
   try {
@@ -21,17 +23,46 @@ async function runMain() {
 
     let exitCode = -1;
 
-    if (buildParameters.providerStrategy === 'local') {
+    // Hot runner path: attempt to use a persistent Unity editor instance
+    if (buildParameters.hotRunnerEnabled) {
+      core.info('[HotRunner] Hot runner mode enabled, attempting hot build...');
+
+      const hotRunnerConfig: HotRunnerConfig = {
+        enabled: true,
+        transport: buildParameters.hotRunnerTransport,
+        host: buildParameters.hotRunnerHost,
+        port: buildParameters.hotRunnerPort,
+        healthCheckInterval: buildParameters.hotRunnerHealthInterval,
+        maxIdleTime: buildParameters.hotRunnerMaxIdle,
+        maxJobsBeforeRecycle: 0, // no automatic recycle by job count
+      };
+
+      const hotRunnerService = new HotRunnerService();
+
+      try {
+        await hotRunnerService.initialize(hotRunnerConfig);
+        const result = await hotRunnerService.submitBuild(buildParameters, (output) => {
+          core.info(output);
+        });
+
+        exitCode = result.exitCode;
+        core.info(`[HotRunner] Build completed with exit code ${exitCode}`);
+        await hotRunnerService.shutdown();
+      } catch (hotRunnerError) {
+        await hotRunnerService.shutdown();
+
+        if (buildParameters.hotRunnerFallbackToCold) {
+          core.warning(
+            `[HotRunner] Hot runner failed: ${(hotRunnerError as Error).message}. Falling back to cold build.`,
+          );
+          exitCode = await runColdBuild(buildParameters, baseImage, workspace, actionFolder);
+        } else {
+          throw hotRunnerError;
+        }
+      }
+    } else if (buildParameters.providerStrategy === 'local') {
       core.info('Building locally');
-      await PlatformSetup.setup(buildParameters, actionFolder);
-      exitCode =
-        process.platform === 'darwin'
-          ? await MacBuilder.run(actionFolder)
-          : await Docker.run(baseImage.toString(), {
-              workspace,
-              actionFolder,
-              ...buildParameters,
-            });
+      exitCode = await runColdBuild(buildParameters, baseImage, workspace, actionFolder);
     } else {
       await Orchestrator.run(buildParameters, baseImage.toString());
       exitCode = 0;
@@ -50,4 +81,28 @@ async function runMain() {
   }
 }
 
+async function runColdBuild(
+  buildParameters: BuildParameters,
+  baseImage: ImageTag,
+  workspace: string,
+  actionFolder: string,
+): Promise<number> {
+  if (buildParameters.providerStrategy === 'local') {
+    core.info('Building locally');
+    await PlatformSetup.setup(buildParameters, actionFolder);
+
+    return process.platform === 'darwin'
+      ? await MacBuilder.run(actionFolder)
+      : await Docker.run(baseImage.toString(), {
+          workspace,
+          actionFolder,
+          ...buildParameters,
+        });
+  } else {
+    await Orchestrator.run(buildParameters, baseImage.toString());
+
+    return 0;
+  }
+}
+
 runMain();

src/model/build-parameters.ts

@@ -54,12 +54,6 @@ class BuildParameters {
   public sshAgent!: string;
   public sshPublicKeysDirectoryPath!: string;
   public providerStrategy!: string;
-  public fallbackProviderStrategy!: string;
-  public runnerCheckEnabled!: boolean;
-  public runnerCheckLabels!: string[];
-  public runnerCheckMinAvailable!: number;
-  public retryOnFallback!: boolean;
-  public providerInitTimeout!: number;
   public gitPrivateToken!: string;
   public awsStackName!: string;
   public awsEndpoint?: string;
@@ -112,6 +106,13 @@ class BuildParameters {
   public cacheUnityInstallationOnMac!: boolean;
   public unityHubVersionOnMac!: string;
   public dockerWorkspacePath!: string;
+  public hotRunnerEnabled!: boolean;
+  public hotRunnerTransport!: 'websocket' | 'grpc' | 'named-pipe';
+  public hotRunnerHost!: string;
+  public hotRunnerPort!: number;
+  public hotRunnerHealthInterval!: number;
+  public hotRunnerMaxIdle!: number;
+  public hotRunnerFallbackToCold!: boolean;
 
   public static shouldUseRetainedWorkspaceMode(buildParameters: BuildParameters) {
     return buildParameters.maxRetainedWorkspaces > 0 && Orchestrator.lockedWorkspace !== ``;
@@ -200,12 +201,6 @@ class BuildParameters {
       containerRegistryRepository: Input.containerRegistryRepository,
       containerRegistryImageVersion: Input.containerRegistryImageVersion,
       providerStrategy: OrchestratorOptions.providerStrategy,
-      fallbackProviderStrategy: OrchestratorOptions.fallbackProviderStrategy,
-      runnerCheckEnabled: OrchestratorOptions.runnerCheckEnabled,
-      runnerCheckLabels: OrchestratorOptions.runnerCheckLabels,
-      runnerCheckMinAvailable: OrchestratorOptions.runnerCheckMinAvailable,
-      retryOnFallback: OrchestratorOptions.retryOnFallback,
-      providerInitTimeout: OrchestratorOptions.providerInitTimeout,
       buildPlatform: OrchestratorOptions.buildPlatform,
       kubeConfig: OrchestratorOptions.kubeConfig,
       containerMemory: OrchestratorOptions.containerMemory,
@@ -254,6 +249,13 @@ class BuildParameters {
       cacheUnityInstallationOnMac: Input.cacheUnityInstallationOnMac,
       unityHubVersionOnMac: Input.unityHubVersionOnMac,
       dockerWorkspacePath: Input.dockerWorkspacePath,
+      hotRunnerEnabled: Input.hotRunnerEnabled,
+      hotRunnerTransport: Input.hotRunnerTransport,
+      hotRunnerHost: Input.hotRunnerHost,
+      hotRunnerPort: Input.hotRunnerPort,
+      hotRunnerHealthInterval: Input.hotRunnerHealthInterval,
+      hotRunnerMaxIdle: Input.hotRunnerMaxIdle,
+      hotRunnerFallbackToCold: Input.hotRunnerFallbackToCold,
     };
   }
 

src/model/input.ts

@@ -282,6 +282,38 @@ class Input {
     return Input.getInput('skipActivation')?.toLowerCase() ?? 'false';
   }
 
+  static get hotRunnerEnabled(): boolean {
+    const input = Input.getInput('hotRunnerEnabled') ?? false;
+
+    return input === 'true';
+  }
+
+  static get hotRunnerTransport(): 'websocket' | 'grpc' | 'named-pipe' {
+    return (Input.getInput('hotRunnerTransport') ?? 'websocket') as 'websocket' | 'grpc' | 'named-pipe';
+  }
+
+  static get hotRunnerHost(): string {
+    return Input.getInput('hotRunnerHost') ?? 'localhost';
+  }
+
+  static get hotRunnerPort(): number {
+    return Number.parseInt(Input.getInput('hotRunnerPort') ?? '9090', 10);
+  }
+
+  static get hotRunnerHealthInterval(): number {
+    return Number.parseInt(Input.getInput('hotRunnerHealthInterval') ?? '30', 10);
+  }
+
+  static get hotRunnerMaxIdle(): number {
+    return Number.parseInt(Input.getInput('hotRunnerMaxIdle') ?? '3600', 10);
+  }
+
+  static get hotRunnerFallbackToCold(): boolean {
+    const input = Input.getInput('hotRunnerFallbackToCold') ?? 'true';
+
+    return input === 'true';
+  }
+
   public static ToEnvVarFormat(input: string) {
     if (input.toUpperCase() === input) {
       return input;

src/model/orchestrator/options/orchestrator-options.ts

@@ -138,32 +138,6 @@ class OrchestratorOptions {
     return provider || 'local';
   }
 
-  static get fallbackProviderStrategy(): string {
-    return OrchestratorOptions.getInput('fallbackProviderStrategy') || '';
-  }
-
-  static get runnerCheckEnabled(): boolean {
-    return OrchestratorOptions.getInput('runnerCheckEnabled') === 'true';
-  }
-
-  static get runnerCheckLabels(): string[] {
-    const labels = OrchestratorOptions.getInput('runnerCheckLabels');
-
-    return labels ? labels.split(',').map((l) => l.trim()) : [];
-  }
-
-  static get runnerCheckMinAvailable(): number {
-    return Number(OrchestratorOptions.getInput('runnerCheckMinAvailable')) || 1;
-  }
-
-  static get retryOnFallback(): boolean {
-    return OrchestratorOptions.getInput('retryOnFallback') === 'true';
-  }
-
-  static get providerInitTimeout(): number {
-    return Number(OrchestratorOptions.getInput('providerInitTimeout')) || 0;
-  }
-
   static get containerCpu(): string {
     return OrchestratorOptions.getInput('containerCpu') || `1024`;
   }

src/model/orchestrator/orchestrator.ts

@@ -20,7 +20,6 @@ import { FollowLogStreamService } from './services/core/follow-log-stream-servic
 import OrchestratorResult from './services/core/orchestrator-result';
 import OrchestratorOptions from './options/orchestrator-options';
 import ResourceTracking from './services/core/resource-tracking';
-import { RunnerAvailabilityService } from './services/core/runner-availability-service';
 
 class Orchestrator {
   public static Provider: ProviderInterface;
@@ -77,42 +76,6 @@ class Orchestrator {
   private static async setupSelectedBuildPlatform() {
     OrchestratorLogger.log(`Orchestrator platform selected ${Orchestrator.buildParameters.providerStrategy}`);
 
-    // Check runner availability and apply fallback if needed
-    if (Orchestrator.buildParameters.runnerCheckEnabled && Orchestrator.buildParameters.fallbackProviderStrategy) {
-      const owner = OrchestratorOptions.githubOwner;
-      const repo = OrchestratorOptions.githubRepoName;
-      const token = Orchestrator.buildParameters.gitPrivateToken || process.env.GITHUB_TOKEN || '';
-
-      OrchestratorLogger.log(
-        `Checking runner availability (labels: [${Orchestrator.buildParameters.runnerCheckLabels.join(', ')}], min: ${
-          Orchestrator.buildParameters.runnerCheckMinAvailable
-        })`,
-      );
-
-      const result = await RunnerAvailabilityService.checkAvailability(
-        owner,
-        repo,
-        token,
-        Orchestrator.buildParameters.runnerCheckLabels,
-        Orchestrator.buildParameters.runnerCheckMinAvailable,
-      );
-
-      OrchestratorLogger.log(
-        `Runner check: ${result.totalRunners} total, ${result.matchingRunners} matching, ${result.idleRunners} idle — ${result.reason}`,
-      );
-
-      if (result.shouldFallback) {
-        const original = Orchestrator.buildParameters.providerStrategy;
-        const fallback = Orchestrator.buildParameters.fallbackProviderStrategy;
-        OrchestratorLogger.log(`Falling back from '${original}' to '${fallback}' — ${result.reason}`);
-        Orchestrator.buildParameters.providerStrategy = fallback;
-        core.setOutput('providerFallbackUsed', 'true');
-        core.setOutput('providerFallbackReason', result.reason);
-      } else {
-        core.setOutput('providerFallbackUsed', 'false');
-      }
-    }
-
     // Detect LocalStack endpoints and handle AWS provider appropriately
     // AWS_FORCE_PROVIDER options:
     //   - 'aws': Force AWS provider (requires LocalStack Pro with ECS support)
@@ -219,30 +182,6 @@ class Orchestrator {
     if (baseImage.includes(`undefined`)) {
       throw new Error(`baseImage is undefined`);
     }
-
-    try {
-      return await Orchestrator.runWithProvider(buildParameters, baseImage);
-    } catch (primaryError: any) {
-      // Retry on fallback provider if enabled and a fallback is configured
-      const fallback = buildParameters.fallbackProviderStrategy;
-      const alreadyOnFallback = buildParameters.providerStrategy === fallback;
-      if (buildParameters.retryOnFallback && fallback && !alreadyOnFallback) {
-        OrchestratorLogger.log(
-          `Primary provider '${buildParameters.providerStrategy}' failed: ${primaryError.message}`,
-        );
-        OrchestratorLogger.log(`Retrying build on fallback provider '${fallback}'...`);
-        buildParameters.providerStrategy = fallback;
-        core.setOutput('providerFallbackUsed', 'true');
-        core.setOutput('providerFallbackReason', `Primary provider failed: ${primaryError.message}`);
-
-        return await Orchestrator.runWithProvider(buildParameters, baseImage);
-      }
-
-      throw primaryError;
-    }
-  }
-
-  private static async runWithProvider(buildParameters: BuildParameters, baseImage: string) {
     await Orchestrator.setup(buildParameters);
 
     // When aws-local mode is enabled, validate AWS CloudFormation templates
@@ -250,10 +189,12 @@ class Orchestrator {
     if (Orchestrator.validateAwsTemplates) {
       await Orchestrator.validateAwsCloudFormationTemplates();
     }
-
-    // Setup workflow with optional init timeout
-    await Orchestrator.setupWorkflowWithTimeout();
-
+    await Orchestrator.Provider.setupWorkflow(
+      Orchestrator.buildParameters.buildGuid,
+      Orchestrator.buildParameters,
+      Orchestrator.buildParameters.branch,
+      Orchestrator.defaultSecrets,
+    );
     try {
       if (buildParameters.maxRetainedWorkspaces > 0) {
         Orchestrator.lockedWorkspace = SharedWorkspaceLocking.NewWorkspaceName();
@@ -334,39 +275,6 @@ class Orchestrator {
     }
   }
 
-  /**
-   * Runs setupWorkflow with an optional timeout. If providerInitTimeout is set and the
-   * provider takes longer than that to initialize, throws an error that triggers
-   * retry-on-fallback (if enabled).
-   */
-  private static async setupWorkflowWithTimeout() {
-    const timeoutSeconds = Orchestrator.buildParameters.providerInitTimeout;
-
-    const setupPromise = Orchestrator.Provider.setupWorkflow(
-      Orchestrator.buildParameters.buildGuid,
-      Orchestrator.buildParameters,
-      Orchestrator.buildParameters.branch,
-      Orchestrator.defaultSecrets,
-    );
-
-    if (timeoutSeconds <= 0) {
-      await setupPromise;
-
-      return;
-    }
-
-    OrchestratorLogger.log(`Provider init timeout: ${timeoutSeconds}s`);
-
-    const timeoutPromise = new Promise<never>((_, reject) => {
-      setTimeout(
-        () => reject(new Error(`Provider initialization timed out after ${timeoutSeconds}s`)),
-        timeoutSeconds * 1000,
-      );
-    });
-
-    await Promise.race([setupPromise, timeoutPromise]);
-  }
-
   private static async updateStatusWithBuildParameters() {
     const content = { ...Orchestrator.buildParameters };
     content.gitPrivateToken = ``;

src/model/orchestrator/runners/README.md

@@ -0,0 +1,5 @@
+# Hot Runner Protocol
+
+Extensible runner registration and persistent Unity editor provider protocol.
+
+See GitHub Issue for full specification.

src/model/orchestrator/services/core/runner-availability-service.test.ts

@@ -1,318 +0,0 @@
-import { RunnerAvailabilityService } from './runner-availability-service';
-
-// Mock @octokit/core
-jest.mock('@octokit/core', () => ({
-  Octokit: jest.fn().mockImplementation(() => ({
-    request: jest.fn(),
-  })),
-}));
-
-jest.mock('./orchestrator-logger', () => ({
-  __esModule: true,
-  default: {
-    log: jest.fn(),
-    logWarning: jest.fn(),
-    error: jest.fn(),
-  },
-}));
-
-import { Octokit } from '@octokit/core';
-
-const MockedOctokit = Octokit as jest.MockedClass<typeof Octokit>;
-
-function createMockRunners(runners: Array<{ name: string; status: string; busy: boolean; labels: string[] }>) {
-  return runners.map((r, i) => ({
-    id: i + 1,
-    name: r.name,
-    status: r.status,
-    busy: r.busy,
-    labels: r.labels.map((l) => ({ name: l })),
-  }));
-}
-
-describe('RunnerAvailabilityService', () => {
-  beforeEach(() => {
-    jest.clearAllMocks();
-  });
-
-  describe('checkAvailability', () => {
-    it('should skip check and not fallback when no token is provided', async () => {
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', '', [], 1);
-      expect(result.shouldFallback).toBe(false);
-      expect(result.reason).toContain('No GitHub token');
-    });
-
-    it('should fallback when no runners are registered', async () => {
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners: [] } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-      expect(result.shouldFallback).toBe(true);
-      expect(result.reason).toContain('No runners registered');
-      expect(result.totalRunners).toBe(0);
-    });
-
-    it('should not fallback when enough idle runners are available', async () => {
-      const runners = createMockRunners([
-        { name: 'runner-1', status: 'online', busy: false, labels: ['self-hosted', 'linux'] },
-        { name: 'runner-2', status: 'online', busy: false, labels: ['self-hosted', 'linux'] },
-      ]);
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-      expect(result.shouldFallback).toBe(false);
-      expect(result.idleRunners).toBe(2);
-      expect(result.totalRunners).toBe(2);
-    });
-
-    it('should fallback when all runners are busy', async () => {
-      const runners = createMockRunners([
-        { name: 'runner-1', status: 'online', busy: true, labels: ['self-hosted'] },
-        { name: 'runner-2', status: 'online', busy: true, labels: ['self-hosted'] },
-      ]);
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-      expect(result.shouldFallback).toBe(true);
-      expect(result.idleRunners).toBe(0);
-      expect(result.matchingRunners).toBe(2);
-    });
-
-    it('should fallback when all runners are offline', async () => {
-      const runners = createMockRunners([
-        { name: 'runner-1', status: 'offline', busy: false, labels: ['self-hosted'] },
-      ]);
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-      expect(result.shouldFallback).toBe(true);
-      expect(result.idleRunners).toBe(0);
-    });
-
-    it('should filter runners by required labels', async () => {
-      const runners = createMockRunners([
-        { name: 'linux-runner', status: 'online', busy: false, labels: ['self-hosted', 'linux'] },
-        { name: 'windows-runner', status: 'online', busy: false, labels: ['self-hosted', 'windows'] },
-      ]);
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability(
-        'owner',
-        'repo',
-        'token',
-        ['self-hosted', 'linux'],
-        1,
-      );
-
-      expect(result.shouldFallback).toBe(false);
-      expect(result.matchingRunners).toBe(1);
-      expect(result.idleRunners).toBe(1);
-      expect(result.totalRunners).toBe(2);
-    });
-
-    it('should fallback when no runners match required labels', async () => {
-      const runners = createMockRunners([
-        { name: 'windows-runner', status: 'online', busy: false, labels: ['self-hosted', 'windows'] },
-      ]);
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability(
-        'owner',
-        'repo',
-        'token',
-        ['self-hosted', 'linux'],
-        1,
-      );
-
-      expect(result.shouldFallback).toBe(true);
-      expect(result.matchingRunners).toBe(0);
-      expect(result.idleRunners).toBe(0);
-    });
-
-    it('should respect minAvailable threshold', async () => {
-      const runners = createMockRunners([{ name: 'runner-1', status: 'online', busy: false, labels: ['self-hosted'] }]);
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      // Need 2, have 1 — should fallback
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 2);
-      expect(result.shouldFallback).toBe(true);
-      expect(result.idleRunners).toBe(1);
-    });
-
-    it('should be case-insensitive for label matching', async () => {
-      const runners = createMockRunners([
-        { name: 'runner-1', status: 'online', busy: false, labels: ['Self-Hosted', 'Linux'] },
-      ]);
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability(
-        'owner',
-        'repo',
-        'token',
-        ['self-hosted', 'linux'],
-        1,
-      );
-      expect(result.shouldFallback).toBe(false);
-      expect(result.matchingRunners).toBe(1);
-    });
-
-    it('should not fallback on API error (fail-open)', async () => {
-      const mockRequest = jest.fn().mockRejectedValue(new Error('403 Forbidden'));
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-      expect(result.shouldFallback).toBe(false);
-      expect(result.reason).toContain('Runner check failed');
-    });
-
-    it('should count only online+idle runners', async () => {
-      const runners = createMockRunners([
-        { name: 'idle', status: 'online', busy: false, labels: ['self-hosted'] },
-        { name: 'busy', status: 'online', busy: true, labels: ['self-hosted'] },
-        { name: 'offline', status: 'offline', busy: false, labels: ['self-hosted'] },
-      ]);
-      const mockRequest = jest.fn().mockResolvedValue({ status: 200, data: { runners } });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-      expect(result.shouldFallback).toBe(false);
-      expect(result.totalRunners).toBe(3);
-      expect(result.matchingRunners).toBe(3);
-      expect(result.idleRunners).toBe(1);
-    });
-  });
-
-  describe('pagination limits', () => {
-    it('should stop paginating after reaching the page limit', async () => {
-      // Return full pages (100 runners each) to force continued pagination
-      let callCount = 0;
-      const mockRequest = jest.fn().mockImplementation(() => {
-        callCount++;
-        const runners = createMockRunners(
-          Array.from({ length: 100 }, (_, i) => ({
-            name: `runner-${callCount}-${i}`,
-            status: 'online' as const,
-            busy: false,
-            labels: ['self-hosted'],
-          })),
-        );
-
-        return Promise.resolve({ status: 200, data: { runners } });
-      });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-
-      // Should have called at most 100 pages (the MAX_PAGINATION_PAGES limit)
-      expect(mockRequest).toHaveBeenCalledTimes(100);
-      // Should still have runners from the pages it did fetch
-      expect(result.totalRunners).toBe(10000);
-      expect(result.shouldFallback).toBe(false);
-    });
-
-    it('should stop paginating on rate limit (HTTP 403)', async () => {
-      let callCount = 0;
-      const mockRequest = jest.fn().mockImplementation(() => {
-        callCount++;
-        if (callCount === 2) {
-          // Octokit throws for non-2xx responses
-          const error: any = new Error('API rate limit exceeded');
-          error.status = 403;
-          error.response = {
-            status: 403,
-            headers: { 'x-ratelimit-reset': String(Math.floor(Date.now() / 1000) + 3600) },
-          };
-
-          return Promise.reject(error);
-        }
-        const runners = createMockRunners(
-          Array.from({ length: 100 }, (_, i) => ({
-            name: `runner-${i}`,
-            status: 'online' as const,
-            busy: false,
-            labels: ['self-hosted'],
-          })),
-        );
-
-        return Promise.resolve({ status: 200, data: { runners } });
-      });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-
-      // Should have stopped at page 2 (rate limited)
-      expect(mockRequest).toHaveBeenCalledTimes(2);
-      // Should use the 100 runners from the first page
-      expect(result.totalRunners).toBe(100);
-      expect(result.shouldFallback).toBe(false);
-    });
-
-    it('should stop paginating on rate limit (HTTP 429)', async () => {
-      let callCount = 0;
-      const mockRequest = jest.fn().mockImplementation(() => {
-        callCount++;
-        if (callCount === 1) {
-          // Octokit throws for non-2xx responses
-          const error: any = new Error('Too Many Requests');
-          error.status = 429;
-          error.response = { status: 429, headers: {} };
-
-          return Promise.reject(error);
-        }
-
-        return Promise.resolve({ status: 200, data: { runners: [] } });
-      });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-
-      // Should have stopped at first page (rate limited immediately)
-      expect(mockRequest).toHaveBeenCalledTimes(1);
-      // No runners found — should fallback
-      expect(result.totalRunners).toBe(0);
-      expect(result.shouldFallback).toBe(true);
-    });
-
-    it('should handle pagination timeout gracefully', async () => {
-      // Mock Date.now to simulate timeout
-      const originalDateNow = Date.now;
-      let callCount = 0;
-
-      const mockRequest = jest.fn().mockImplementation(() => {
-        callCount++;
-        // After first call, advance time past the timeout
-        if (callCount >= 2) {
-          Date.now = jest.fn(() => originalDateNow() + 31_000);
-        }
-        const runners = createMockRunners(
-          Array.from({ length: 100 }, (_, i) => ({
-            name: `runner-${callCount}-${i}`,
-            status: 'online' as const,
-            busy: false,
-            labels: ['self-hosted'],
-          })),
-        );
-
-        return Promise.resolve({ status: 200, data: { runners } });
-      });
-      MockedOctokit.mockImplementation(() => ({ request: mockRequest } as any));
-
-      const result = await RunnerAvailabilityService.checkAvailability('owner', 'repo', 'token', [], 1);
-
-      // Should have stopped after timeout was detected (2 pages: first succeeds, second triggers timeout check)
-      expect(mockRequest.mock.calls.length).toBeLessThanOrEqual(3);
-      // Should have runners from pages fetched before timeout
-      expect(result.totalRunners).toBeGreaterThan(0);
-
-      // Restore
-      Date.now = originalDateNow;
-    });
-  });
-});

src/model/orchestrator/services/core/runner-availability-service.ts

@@ -1,205 +0,0 @@
-import { Octokit } from '@octokit/core';
-import OrchestratorLogger from './orchestrator-logger';
-
-interface GitHubRunner {
-  id: number;
-  name: string;
-  status: 'online' | 'offline';
-  busy: boolean;
-  labels: Array<{ name: string }>;
-}
-
-interface RunnerCheckResult {
-  shouldFallback: boolean;
-  reason: string;
-  totalRunners: number;
-  matchingRunners: number;
-  idleRunners: number;
-}
-
-/**
- * Maximum number of pages to fetch when paginating through GitHub API results.
- * 100 pages * 100 per page = 10,000 runners maximum.
- */
-const MAX_PAGINATION_PAGES = 100;
-
-/**
- * Total timeout in milliseconds for the pagination loop.
- * Prevents indefinite API calls if GitHub is slow or pagination is unexpectedly deep.
- */
-const PAGINATION_TIMEOUT_MS = 30_000;
-
-/**
- * Checks GitHub Actions runner availability to support automatic provider fallback.
- *
- * When a user configures `runnerCheckEnabled: true` with a `fallbackProviderStrategy`,
- * this service queries the GitHub API for runner status before the build starts.
- * If insufficient runners are available, the orchestrator routes to the fallback provider.
- */
-export class RunnerAvailabilityService {
-  /**
-   * Check if enough runners are available to handle the build.
-   *
-   * @param owner - GitHub repository owner
-   * @param repo - GitHub repository name
-   * @param token - GitHub token with repo/actions scope
-   * @param requiredLabels - Labels runners must have (empty = any runner)
-   * @param minAvailable - Minimum idle runners required
-   * @returns RunnerCheckResult with decision and diagnostics
-   */
-  static async checkAvailability(
-    owner: string,
-    repo: string,
-    token: string,
-    requiredLabels: string[],
-    minAvailable: number,
-  ): Promise<RunnerCheckResult> {
-    if (!token) {
-      return {
-        shouldFallback: false,
-        reason: 'No GitHub token available — skipping runner check',
-        totalRunners: 0,
-        matchingRunners: 0,
-        idleRunners: 0,
-      };
-    }
-
-    try {
-      const octokit = new Octokit({ auth: token });
-
-      // Fetch all runners for the repository
-      const runners = await RunnerAvailabilityService.fetchRunners(octokit, owner, repo);
-
-      if (runners.length === 0) {
-        return {
-          shouldFallback: true,
-          reason: 'No runners registered for this repository',
-          totalRunners: 0,
-          matchingRunners: 0,
-          idleRunners: 0,
-        };
-      }
-
-      // Filter by required labels
-      const matching = RunnerAvailabilityService.filterByLabels(runners, requiredLabels);
-
-      // Count idle (online + not busy)
-      const idle = matching.filter((r) => r.status === 'online' && !r.busy);
-
-      const result: RunnerCheckResult = {
-        shouldFallback: idle.length < minAvailable,
-        reason:
-          idle.length >= minAvailable
-            ? `${idle.length} idle runner(s) available (need ${minAvailable})`
-            : `Only ${idle.length} idle runner(s) available, need ${minAvailable}`,
-        totalRunners: runners.length,
-        matchingRunners: matching.length,
-        idleRunners: idle.length,
-      };
-
-      return result;
-    } catch (error: any) {
-      // If the API call fails (permissions, rate limit, etc.), don't block the build
-      OrchestratorLogger.log(`Runner availability check failed: ${error.message}`);
-
-      return {
-        shouldFallback: false,
-        reason: `Runner check failed (${error.message}) — proceeding with primary provider`,
-        totalRunners: 0,
-        matchingRunners: 0,
-        idleRunners: 0,
-      };
-    }
-  }
-
-  /**
-   * Fetch all runners for a repository, handling pagination.
-   *
-   * Includes defensive limits:
-   * - Maximum page count (MAX_PAGINATION_PAGES) to prevent infinite loops
-   * - Total timeout (PAGINATION_TIMEOUT_MS) to prevent indefinite API calls
-   * - Rate-limit detection (HTTP 403/429 with X-RateLimit-Remaining header)
-   */
-  private static async fetchRunners(octokit: Octokit, owner: string, repo: string): Promise<GitHubRunner[]> {
-    const allRunners: GitHubRunner[] = [];
-    let page = 1;
-    const perPage = 100;
-    const startTime = Date.now();
-
-    while (page <= MAX_PAGINATION_PAGES) {
-      // Check total timeout
-      if (Date.now() - startTime > PAGINATION_TIMEOUT_MS) {
-        OrchestratorLogger.logWarning(
-          `[RunnerAvailability] Pagination timeout reached after ${page - 1} pages and ${Date.now() - startTime}ms. ` +
-            `Using ${allRunners.length} runners found so far.`,
-        );
-        break;
-      }
-
-      let response: any;
-      try {
-        response = await octokit.request('GET /repos/{owner}/{repo}/actions/runners', {
-          owner,
-          repo,
-          per_page: perPage,
-          page,
-        });
-      } catch (requestError: any) {
-        // Octokit throws for non-2xx responses. Check if this is a rate limit error.
-        const status = requestError.status ?? requestError.response?.status;
-        if (status === 403 || status === 429) {
-          const resetTime =
-            requestError.response?.headers?.['x-ratelimit-reset'] ?? requestError.headers?.['x-ratelimit-reset'];
-          const resetMessage = resetTime
-            ? ` Resets at ${new Date(Number.parseInt(String(resetTime), 10) * 1000).toISOString()}`
-            : '';
-          OrchestratorLogger.logWarning(
-            `[RunnerAvailability] GitHub API rate limit reached (HTTP ${status}).${resetMessage} ` +
-              `Using ${allRunners.length} runners found so far.`,
-          );
-          break;
-        }
-        // Re-throw non-rate-limit errors to be handled by the outer catch
-        throw requestError;
-      }
-
-      const runners = (response.data.runners || []) as GitHubRunner[];
-      allRunners.push(...runners);
-
-      if (runners.length < perPage) break;
-      page++;
-    }
-
-    if (page > MAX_PAGINATION_PAGES) {
-      OrchestratorLogger.logWarning(
-        `[RunnerAvailability] Maximum pagination limit reached (${MAX_PAGINATION_PAGES} pages). ` +
-          `Using ${allRunners.length} runners found so far.`,
-      );
-    }
-
-    if (allRunners.length === 0) {
-      OrchestratorLogger.log(
-        '[RunnerAvailability] No runners found. Possible causes: ' +
-          'wrong token permissions (needs repo or actions scope), ' +
-          'no self-hosted runners registered, ' +
-          'or runners are registered at the organization level instead of the repository.',
-      );
-    }
-
-    return allRunners;
-  }
-
-  /**
-   * Filter runners by required labels. A runner matches if it has ALL required labels.
-   * If requiredLabels is empty, all runners match.
-   */
-  private static filterByLabels(runners: GitHubRunner[], requiredLabels: string[]): GitHubRunner[] {
-    if (requiredLabels.length === 0) return runners;
-
-    return runners.filter((runner) => {
-      const runnerLabelNames = runner.labels.map((l) => l.name.toLowerCase());
-
-      return requiredLabels.every((required) => runnerLabelNames.includes(required.toLowerCase()));
-    });
-  }
-}

src/model/orchestrator/services/hot-runner/hot-runner-dispatcher.ts

@@ -0,0 +1,159 @@
+import OrchestratorLogger from '../core/orchestrator-logger';
+import { HotRunnerRegistry } from './hot-runner-registry';
+import { HotRunnerJobRequest, HotRunnerJobResult, HotRunnerStatus, HotRunnerTransport } from './hot-runner-types';
+
+const POLL_INTERVAL_MS = 1000;
+
+// eslint-disable-next-line no-unused-vars
+export type OutputCallback = (output: string) => void;
+
+export class HotRunnerDispatcher {
+  private transports: Map<string, HotRunnerTransport>;
+
+  constructor(transports: Map<string, HotRunnerTransport>) {
+    this.transports = transports;
+  }
+
+  /**
+   * Dispatch a job to an available hot runner matching the request's build target.
+   * If no runner is immediately available, waits up to the request timeout.
+   * Returns the job result, or throws if no runner becomes available in time.
+   */
+  async dispatchJob(
+    request: HotRunnerJobRequest,
+    registry: HotRunnerRegistry,
+    unityVersion: string,
+    onOutput?: OutputCallback,
+  ): Promise<HotRunnerJobResult> {
+    OrchestratorLogger.log(`[HotRunner] Dispatching job ${request.jobId} (target: ${request.buildTarget})`);
+
+    // Find or wait for an available runner
+    let runner = registry.findAvailableRunner({
+      unityVersion,
+      platform: request.buildTarget,
+    });
+
+    if (!runner) {
+      OrchestratorLogger.log(
+        `[HotRunner] No idle runner available for ${unityVersion}/${request.buildTarget}, waiting...`,
+      );
+      runner = await this.waitForRunner({ unityVersion, platform: request.buildTarget }, request.timeout, registry);
+    }
+
+    // Mark runner as busy
+    registry.updateRunner(runner.id, {
+      state: 'busy',
+      currentJob: request.jobId,
+    });
+
+    const transport = this.transports.get(runner.id);
+    if (!transport) {
+      registry.updateRunner(runner.id, { state: 'idle', currentJob: undefined });
+      throw new Error(`[HotRunner] No transport available for runner ${runner.id}`);
+    }
+
+    OrchestratorLogger.log(`[HotRunner] Sending job ${request.jobId} to runner ${runner.id}`);
+
+    const startTime = Date.now();
+
+    try {
+      const result = await this.executeWithTimeout(transport, request);
+
+      const duration = Date.now() - startTime;
+      OrchestratorLogger.log(
+        `[HotRunner] Job ${request.jobId} completed on runner ${runner.id} in ${duration}ms (exit: ${result.exitCode})`,
+      );
+
+      if (onOutput && result.output) {
+        onOutput(result.output);
+      }
+
+      // Mark runner as idle and increment job count
+      const currentStatus = registry.getRunner(runner.id);
+      registry.updateRunner(runner.id, {
+        state: 'idle',
+        currentJob: undefined,
+        lastJobCompleted: request.jobId,
+        jobsCompleted: (currentStatus?.jobsCompleted ?? 0) + 1,
+      });
+
+      return result;
+    } catch (error: any) {
+      OrchestratorLogger.logWarning(`[HotRunner] Job ${request.jobId} failed on runner ${runner.id}: ${error.message}`);
+
+      // Mark runner as idle despite failure -- the health monitor will recycle if needed
+      registry.updateRunner(runner.id, {
+        state: 'idle',
+        currentJob: undefined,
+      });
+
+      throw error;
+    }
+  }
+
+  /**
+   * Wait for an available runner matching the requirements.
+   * Polls the registry at a fixed interval until one becomes available or timeout expires.
+   */
+  async waitForRunner(
+    requirements: { unityVersion: string; platform: string },
+    timeoutMs: number,
+    registry: HotRunnerRegistry,
+  ): Promise<HotRunnerStatus> {
+    const deadline = Date.now() + timeoutMs;
+
+    while (Date.now() < deadline) {
+      const runner = registry.findAvailableRunner(requirements);
+      if (runner) {
+        OrchestratorLogger.log(`[HotRunner] Runner ${runner.id} became available`);
+
+        return runner;
+      }
+
+      await this.sleep(Math.min(POLL_INTERVAL_MS, deadline - Date.now()));
+    }
+
+    throw new Error(
+      `[HotRunner] Timed out waiting for available runner (${requirements.unityVersion}/${requirements.platform}) after ${timeoutMs}ms`,
+    );
+  }
+
+  /**
+   * Execute a job on a transport with a timeout guard.
+   * On timeout, disconnects the transport to release the connection
+   * and prevent the orphaned sendJob promise from holding resources.
+   */
+  private async executeWithTimeout(
+    transport: HotRunnerTransport,
+    request: HotRunnerJobRequest,
+  ): Promise<HotRunnerJobResult> {
+    const TIMEOUT_SENTINEL = Symbol('timeout');
+
+    const timeoutPromise = new Promise<typeof TIMEOUT_SENTINEL>((resolve) => {
+      setTimeout(() => {
+        resolve(TIMEOUT_SENTINEL);
+      }, request.timeout);
+    });
+
+    const result = await Promise.race([transport.sendJob(request), timeoutPromise]);
+
+    if (result === TIMEOUT_SENTINEL) {
+      // Disconnect the transport to clean up the orphaned sendJob call
+      try {
+        await transport.disconnect();
+      } catch (disconnectError: any) {
+        OrchestratorLogger.logWarning(
+          `[HotRunner] Error disconnecting transport after timeout for job ${request.jobId}: ${disconnectError.message}`,
+        );
+      }
+
+      throw new Error(`[HotRunner] Job ${request.jobId} timed out after ${request.timeout}ms`);
+    }
+
+    return result;
+  }
+
+  private sleep(ms: number): Promise<void> {
+    return new Promise((resolve) => setTimeout(resolve, ms));
+  }
+}

src/model/orchestrator/services/hot-runner/hot-runner-health-monitor.ts

@@ -0,0 +1,186 @@
+import OrchestratorLogger from '../core/orchestrator-logger';
+import { HotRunnerRegistry } from './hot-runner-registry';
+import { HotRunnerTransport } from './hot-runner-types';
+
+export class HotRunnerHealthMonitor {
+  private intervalHandle: ReturnType<typeof setInterval> | undefined;
+  private registry: HotRunnerRegistry | undefined;
+  private transports: Map<string, HotRunnerTransport> = new Map();
+
+  /**
+   * Start periodic health monitoring for all registered runners.
+   */
+  startMonitoring(registry: HotRunnerRegistry, interval: number, transports: Map<string, HotRunnerTransport>): void {
+    if (this.intervalHandle) {
+      this.stopMonitoring();
+    }
+
+    this.registry = registry;
+    this.transports = transports;
+
+    OrchestratorLogger.log(`[HotRunner] Starting health monitoring (interval: ${interval}s)`);
+
+    this.intervalHandle = setInterval(() => {
+      this.runHealthChecks().catch((error: any) => {
+        OrchestratorLogger.logWarning(`[HotRunner] Health check cycle failed: ${error.message}`);
+      });
+    }, interval * 1000);
+  }
+
+  /**
+   * Stop periodic health monitoring.
+   */
+  stopMonitoring(): void {
+    if (this.intervalHandle) {
+      clearInterval(this.intervalHandle);
+      this.intervalHandle = undefined;
+      OrchestratorLogger.log(`[HotRunner] Health monitoring stopped`);
+    }
+  }
+
+  /**
+   * Check health of a specific runner by ID. Returns true if healthy.
+   */
+  async checkHealth(runnerId: string): Promise<boolean> {
+    if (!this.registry) {
+      return false;
+    }
+
+    const transport = this.transports.get(runnerId);
+    if (!transport) {
+      OrchestratorLogger.logWarning(`[HotRunner] No transport for runner ${runnerId}`);
+      this.registry.updateRunner(runnerId, {
+        state: 'unhealthy',
+        lastHealthCheck: new Date().toISOString(),
+      });
+
+      return false;
+    }
+
+    try {
+      const healthy = await transport.healthCheck();
+      if (healthy) {
+        const status = await transport.getStatus();
+        this.registry.updateRunner(runnerId, {
+          lastHealthCheck: new Date().toISOString(),
+          memoryUsageMB: status.memoryUsageMB,
+          uptime: status.uptime,
+          libraryHash: status.libraryHash,
+        });
+
+        return true;
+      }
+
+      OrchestratorLogger.logWarning(`[HotRunner] Runner ${runnerId} health check returned false`);
+      this.registry.updateRunner(runnerId, {
+        state: 'unhealthy',
+        lastHealthCheck: new Date().toISOString(),
+      });
+
+      return false;
+    } catch (error: any) {
+      OrchestratorLogger.logWarning(`[HotRunner] Runner ${runnerId} health check failed: ${error.message}`);
+      this.registry.updateRunner(runnerId, {
+        state: 'unhealthy',
+        lastHealthCheck: new Date().toISOString(),
+      });
+
+      return false;
+    }
+  }
+
+  /**
+   * Mark an unhealthy runner for cleanup and disconnect its transport.
+   */
+  async recycleUnhealthyRunner(runnerId: string): Promise<void> {
+    if (!this.registry) {
+      return;
+    }
+
+    OrchestratorLogger.log(`[HotRunner] Recycling unhealthy runner ${runnerId}`);
+    this.registry.updateRunner(runnerId, { state: 'stopping' });
+
+    const transport = this.transports.get(runnerId);
+    if (transport) {
+      try {
+        await transport.disconnect();
+      } catch (error: any) {
+        OrchestratorLogger.logWarning(`[HotRunner] Error disconnecting runner ${runnerId}: ${error.message}`);
+      }
+      this.transports.delete(runnerId);
+    }
+
+    this.registry.unregisterRunner(runnerId);
+    OrchestratorLogger.log(`[HotRunner] Runner ${runnerId} recycled and removed`);
+  }
+
+  /**
+   * Recycle a runner that has been idle longer than the maximum idle time.
+   */
+  async recycleIdleRunner(runnerId: string, maxIdleTime: number): Promise<void> {
+    if (!this.registry) {
+      return;
+    }
+
+    const runner = this.registry.getRunner(runnerId);
+    if (!runner || runner.state !== 'idle') {
+      return;
+    }
+
+    const lastCheckTime = new Date(runner.lastHealthCheck).getTime();
+    const now = Date.now();
+    const idleSeconds = (now - lastCheckTime) / 1000;
+
+    if (idleSeconds >= maxIdleTime) {
+      OrchestratorLogger.log(
+        `[HotRunner] Runner ${runnerId} idle for ${Math.floor(idleSeconds)}s (max: ${maxIdleTime}s), recycling`,
+      );
+      await this.recycleUnhealthyRunner(runnerId);
+    }
+  }
+
+  /**
+   * Run health checks and idle-recycle checks for all registered runners.
+   */
+  private async runHealthChecks(): Promise<void> {
+    if (!this.registry) {
+      return;
+    }
+
+    const runners = this.registry.listRunners();
+
+    for (const runner of runners) {
+      if (runner.state === 'stopping') {
+        continue;
+      }
+
+      const healthy = await this.checkHealth(runner.id);
+
+      if (!healthy && runner.state !== 'starting') {
+        await this.recycleUnhealthyRunner(runner.id);
+        continue;
+      }
+
+      // Check for idle timeout
+      const config = this.registry.getConfig(runner.id);
+      if (config && runner.state === 'idle') {
+        await this.recycleIdleRunner(runner.id, config.maxIdleTime);
+      }
+
+      // Check for max jobs before recycle
+      if (config && config.maxJobsBeforeRecycle > 0 && runner.jobsCompleted >= config.maxJobsBeforeRecycle) {
+        OrchestratorLogger.log(
+          `[HotRunner] Runner ${runner.id} reached max jobs (${runner.jobsCompleted}/${config.maxJobsBeforeRecycle}), recycling`,
+        );
+        await this.recycleUnhealthyRunner(runner.id);
+      }
+    }
+  }
+
+  /**
+   * Whether health monitoring is currently active.
+   */
+  get isMonitoring(): boolean {
+    return this.intervalHandle !== undefined;
+  }
+}

src/model/orchestrator/services/hot-runner/hot-runner-registry.ts

@@ -0,0 +1,315 @@
+import fs from 'node:fs';
+import path from 'node:path';
+import { customAlphabet } from 'nanoid';
+import OrchestratorLogger from '../core/orchestrator-logger';
+import { HotRunnerConfig, HotRunnerStatus } from './hot-runner-types';
+
+const generateId = customAlphabet('abcdefghijklmnopqrstuvwxyz0123456789', 12);
+
+const PERSISTENCE_FILENAME = 'hot-runners.json';
+
+const VALID_RUNNER_STATES: ReadonlySet<string> = new Set(['idle', 'busy', 'starting', 'stopping', 'unhealthy']);
+
+export interface HotRunnerFilter {
+  platform?: string;
+  state?: string;
+  unityVersion?: string;
+}
+
+/**
+ * Validate that a restored runner entry has all required fields with correct types.
+ * Returns true if the entry is a valid HotRunnerStatus, false otherwise.
+ */
+function isValidRunnerStatus(entry: unknown): entry is HotRunnerStatus {
+  if (typeof entry !== 'object' || entry === null) {
+    return false;
+  }
+
+  const record = entry as Record<string, unknown>;
+
+  return (
+    typeof record.id === 'string' &&
+    record.id.length > 0 &&
+    typeof record.state === 'string' &&
+    VALID_RUNNER_STATES.has(record.state) &&
+    typeof record.unityVersion === 'string' &&
+    typeof record.platform === 'string' &&
+    typeof record.uptime === 'number' &&
+    typeof record.jobsCompleted === 'number' &&
+    typeof record.lastHealthCheck === 'string' &&
+    typeof record.memoryUsageMB === 'number'
+  );
+}
+
+/**
+ * Validate that a restored config entry has all required fields with correct types.
+ * Returns true if the entry is a valid HotRunnerConfig, false otherwise.
+ */
+function isValidRunnerConfig(entry: unknown): entry is HotRunnerConfig {
+  if (typeof entry !== 'object' || entry === null) {
+    return false;
+  }
+
+  const record = entry as Record<string, unknown>;
+
+  return (
+    typeof record.enabled === 'boolean' &&
+    typeof record.transport === 'string' &&
+    ['websocket', 'grpc', 'named-pipe'].includes(record.transport) &&
+    typeof record.host === 'string' &&
+    typeof record.port === 'number' &&
+    typeof record.healthCheckInterval === 'number' &&
+    typeof record.maxIdleTime === 'number' &&
+    typeof record.maxJobsBeforeRecycle === 'number'
+  );
+}
+
+export class HotRunnerRegistry {
+  private runners: Map<string, HotRunnerStatus> = new Map();
+  private configs: Map<string, HotRunnerConfig> = new Map();
+  private persistencePath: string;
+
+  constructor(persistenceDirectory?: string) {
+    this.persistencePath = persistenceDirectory ? path.join(persistenceDirectory, PERSISTENCE_FILENAME) : '';
+  }
+
+  /**
+   * Register a new hot runner. Returns the generated runner ID.
+   */
+  registerRunner(config: HotRunnerConfig): string {
+    const id = `hr-${generateId()}`;
+
+    const status: HotRunnerStatus = {
+      id,
+      state: 'starting',
+      unityVersion: config.unityVersion ?? 'unknown',
+      platform: config.platform ?? 'unknown',
+      uptime: 0,
+      jobsCompleted: 0,
+      lastHealthCheck: new Date().toISOString(),
+      memoryUsageMB: 0,
+    };
+
+    this.runners.set(id, status);
+    this.configs.set(id, config);
+    OrchestratorLogger.log(`[HotRunner] Registered runner ${id} (${status.unityVersion}/${status.platform})`);
+
+    this.persist();
+
+    return id;
+  }
+
+  /**
+   * Remove a runner from the registry.
+   */
+  unregisterRunner(id: string): void {
+    const existed = this.runners.delete(id);
+    this.configs.delete(id);
+
+    if (existed) {
+      OrchestratorLogger.log(`[HotRunner] Unregistered runner ${id}`);
+      this.persist();
+    }
+  }
+
+  /**
+   * Get a runner's current status by ID.
+   */
+  getRunner(id: string): HotRunnerStatus | undefined {
+    return this.runners.get(id);
+  }
+
+  /**
+   * Get a runner's config by ID.
+   */
+  getConfig(id: string): HotRunnerConfig | undefined {
+    return this.configs.get(id);
+  }
+
+  /**
+   * List all runners, optionally filtered by platform, state, or Unity version.
+   */
+  listRunners(filter?: HotRunnerFilter): HotRunnerStatus[] {
+    let results = [...this.runners.values()];
+
+    if (filter?.platform) {
+      results = results.filter((runner) => runner.platform === filter.platform);
+    }
+
+    if (filter?.state) {
+      results = results.filter((runner) => runner.state === filter.state);
+    }
+
+    if (filter?.unityVersion) {
+      results = results.filter((runner) => runner.unityVersion === filter.unityVersion);
+    }
+
+    return results;
+  }
+
+  /**
+   * Find an idle runner matching the given Unity version and platform requirements.
+   */
+  findAvailableRunner(requirements: { unityVersion: string; platform: string }): HotRunnerStatus | undefined {
+    return this.listRunners({
+      state: 'idle',
+      unityVersion: requirements.unityVersion,
+      platform: requirements.platform,
+    })[0];
+  }
+
+  /**
+   * Update a runner's status fields. Merges partial updates into existing status.
+   */
+  updateRunner(id: string, update: Partial<HotRunnerStatus>): void {
+    const existing = this.runners.get(id);
+    if (!existing) {
+      return;
+    }
+
+    this.runners.set(id, { ...existing, ...update, id });
+    this.persist();
+  }
+
+  /**
+   * Get the total number of registered runners.
+   */
+  get size(): number {
+    return this.runners.size;
+  }
+
+  /**
+   * Validate all runners in the registry and reset invalid ones to 'unhealthy'.
+   * Returns the number of runners that were repaired.
+   */
+  validateAndRepair(): number {
+    let repaired = 0;
+
+    for (const [id, status] of this.runners) {
+      // Cast to unknown to bypass the type guard narrowing to 'never',
+      // since the Map is typed as HotRunnerStatus but entries may have
+      // been corrupted via direct deserialization or unsafe casts.
+      const entry = status as unknown as Record<string, unknown>;
+
+      if (!isValidRunnerStatus(entry)) {
+        OrchestratorLogger.logWarning(`[HotRunner] Runner ${id} has invalid state, marking as unhealthy`);
+        this.runners.set(id, {
+          id,
+          state: 'unhealthy',
+          unityVersion: typeof entry.unityVersion === 'string' ? entry.unityVersion : 'unknown',
+          platform: typeof entry.platform === 'string' ? entry.platform : 'unknown',
+          uptime: typeof entry.uptime === 'number' ? entry.uptime : 0,
+          jobsCompleted: typeof entry.jobsCompleted === 'number' ? entry.jobsCompleted : 0,
+          lastHealthCheck: typeof entry.lastHealthCheck === 'string' ? entry.lastHealthCheck : new Date().toISOString(),
+          memoryUsageMB: typeof entry.memoryUsageMB === 'number' ? entry.memoryUsageMB : 0,
+        });
+        repaired++;
+      }
+    }
+
+    if (repaired > 0) {
+      this.persist();
+    }
+
+    return repaired;
+  }
+
+  /**
+   * Persist current registry state to disk for crash recovery.
+   * Validates data before writing to prevent persisting corrupt state.
+   */
+  private persist(): void {
+    if (!this.persistencePath) {
+      return;
+    }
+
+    try {
+      // Validate data before persisting
+      for (const [id, status] of this.runners) {
+        if (!isValidRunnerStatus(status)) {
+          OrchestratorLogger.logWarning(`[HotRunner] Skipping persistence -- runner ${id} has invalid state`);
+
+          return;
+        }
+      }
+
+      const data = {
+        runners: Object.fromEntries(this.runners),
+        configs: Object.fromEntries(this.configs),
+      };
+      const directory = path.dirname(this.persistencePath);
+      if (!fs.existsSync(directory)) {
+        fs.mkdirSync(directory, { recursive: true });
+      }
+
+      fs.writeFileSync(this.persistencePath, JSON.stringify(data, undefined, 2));
+    } catch (error: any) {
+      OrchestratorLogger.logWarning(`[HotRunner] Failed to persist registry: ${error.message}`);
+    }
+  }
+
+  /**
+   * Load registry state from disk. Returns the number of runners restored.
+   * Validates each restored entry and discards corrupt entries with warnings.
+   * If the persistence file itself is corrupt (invalid JSON), starts with
+   * an empty registry.
+   */
+  loadFromDisk(): number {
+    if (!this.persistencePath || !fs.existsSync(this.persistencePath)) {
+      return 0;
+    }
+
+    let data: any;
+
+    try {
+      const raw = fs.readFileSync(this.persistencePath, 'utf8');
+      data = JSON.parse(raw);
+    } catch (error: any) {
+      OrchestratorLogger.logWarning(
+        `[HotRunner] Persistence file is corrupt, starting with empty registry: ${error.message}`,
+      );
+
+      return 0;
+    }
+
+    if (typeof data !== 'object' || data === null) {
+      OrchestratorLogger.logWarning('[HotRunner] Persistence file has invalid structure, starting with empty registry');
+
+      return 0;
+    }
+
+    let discarded = 0;
+
+    if (data.runners && typeof data.runners === 'object') {
+      for (const [id, status] of Object.entries(data.runners)) {
+        if (isValidRunnerStatus(status)) {
+          this.runners.set(id, status);
+        } else {
+          OrchestratorLogger.logWarning(`[HotRunner] Discarding invalid runner entry '${id}' from persistence file`);
+          discarded++;
+        }
+      }
+    }
+
+    if (data.configs && typeof data.configs === 'object') {
+      for (const [id, config] of Object.entries(data.configs)) {
+        // Only restore configs for runners that were successfully restored
+        if (this.runners.has(id)) {
+          if (isValidRunnerConfig(config)) {
+            this.configs.set(id, config);
+          } else {
+            OrchestratorLogger.logWarning(`[HotRunner] Discarding invalid config entry '${id}' from persistence file`);
+          }
+        }
+      }
+    }
+
+    if (discarded > 0) {
+      OrchestratorLogger.logWarning(`[HotRunner] Discarded ${discarded} invalid runner(s) from persistence file`);
+    }
+
+    OrchestratorLogger.log(`[HotRunner] Restored ${this.runners.size} runner(s) from disk`);
+
+    return this.runners.size;
+  }
+}

src/model/orchestrator/services/hot-runner/hot-runner-service.ts

@@ -0,0 +1,166 @@
+import OrchestratorLogger from '../core/orchestrator-logger';
+import { HotRunnerRegistry } from './hot-runner-registry';
+import { HotRunnerHealthMonitor } from './hot-runner-health-monitor';
+import { HotRunnerDispatcher, OutputCallback } from './hot-runner-dispatcher';
+import {
+  HotRunnerConfig,
+  HotRunnerJobRequest,
+  HotRunnerJobResult,
+  HotRunnerStatus,
+  HotRunnerTransport,
+} from './hot-runner-types';
+import BuildParameters from '../../../build-parameters';
+
+export class HotRunnerService {
+  private registry: HotRunnerRegistry;
+  private healthMonitor: HotRunnerHealthMonitor;
+  private dispatcher: HotRunnerDispatcher;
+  private transports: Map<string, HotRunnerTransport> = new Map();
+  private config: HotRunnerConfig | undefined;
+
+  constructor(persistenceDirectory?: string) {
+    this.registry = new HotRunnerRegistry(persistenceDirectory);
+    this.healthMonitor = new HotRunnerHealthMonitor();
+    this.dispatcher = new HotRunnerDispatcher(this.transports);
+  }
+
+  /**
+   * Initialize the hot runner service: load persisted state, start health monitoring.
+   */
+  async initialize(config: HotRunnerConfig): Promise<void> {
+    this.config = config;
+
+    OrchestratorLogger.log(
+      `[HotRunner] Initializing service (transport: ${config.transport}, ${config.host}:${config.port})`,
+    );
+
+    // Attempt to restore previously registered runners from disk
+    const restored = this.registry.loadFromDisk();
+    if (restored > 0) {
+      OrchestratorLogger.log(`[HotRunner] Restored ${restored} runner(s) from persistence`);
+    }
+
+    // Start health monitoring
+    this.healthMonitor.startMonitoring(this.registry, config.healthCheckInterval, this.transports);
+
+    OrchestratorLogger.log(`[HotRunner] Service initialized`);
+  }
+
+  /**
+   * Register a runner with a transport implementation.
+   * Returns the runner ID.
+   */
+  registerRunner(config: HotRunnerConfig, transport: HotRunnerTransport): string {
+    const id = this.registry.registerRunner(config);
+    this.transports.set(id, transport);
+
+    return id;
+  }
+
+  /**
+   * Submit a build job to an available hot runner.
+   * Converts BuildParameters to a HotRunnerJobRequest and dispatches.
+   */
+  async submitBuild(params: BuildParameters, onOutput?: OutputCallback): Promise<HotRunnerJobResult> {
+    const request: HotRunnerJobRequest = {
+      jobId: params.buildGuid || `build-${Date.now()}`,
+      buildMethod: params.buildMethod || undefined,
+      buildTarget: params.targetPlatform,
+      buildPath: params.buildPath,
+      customParameters: params.customParameters ? this.parseCustomParameters(params.customParameters) : undefined,
+      timeout: 30 * 60 * 1000, // 30 minutes default
+    };
+
+    OrchestratorLogger.log(`[HotRunner] Submitting build: ${request.jobId} (target: ${request.buildTarget})`);
+
+    return this.dispatcher.dispatchJob(request, this.registry, params.editorVersion, onOutput);
+  }
+
+  /**
+   * Submit a test job to an available hot runner.
+   * Converts BuildParameters and optional suite config to a test-mode HotRunnerJobRequest.
+   */
+  async submitTest(
+    params: BuildParameters,
+    suiteConfig?: { testMode?: 'editmode' | 'playmode'; testSuitePath?: string },
+    onOutput?: OutputCallback,
+  ): Promise<HotRunnerJobResult> {
+    const request: HotRunnerJobRequest = {
+      jobId: params.buildGuid || `test-${Date.now()}`,
+      buildTarget: params.targetPlatform,
+      customParameters: params.customParameters ? this.parseCustomParameters(params.customParameters) : undefined,
+      timeout: 30 * 60 * 1000, // 30 minutes default
+      testMode: suiteConfig?.testMode ?? 'editmode',
+      testSuitePath: suiteConfig?.testSuitePath,
+    };
+
+    OrchestratorLogger.log(`[HotRunner] Submitting test: ${request.jobId} (mode: ${request.testMode})`);
+
+    return this.dispatcher.dispatchJob(request, this.registry, params.editorVersion, onOutput);
+  }
+
+  /**
+   * Shut down the service: stop health monitoring, disconnect all transports,
+   * and unregister all runners.
+   */
+  async shutdown(): Promise<void> {
+    OrchestratorLogger.log(`[HotRunner] Shutting down service`);
+
+    this.healthMonitor.stopMonitoring();
+
+    const disconnectPromises: Promise<void>[] = [];
+    for (const [id, transport] of this.transports.entries()) {
+      disconnectPromises.push(
+        transport.disconnect().catch((error: any) => {
+          OrchestratorLogger.logWarning(`[HotRunner] Error disconnecting runner ${id}: ${error.message}`);
+        }),
+      );
+    }
+    await Promise.all(disconnectPromises);
+
+    this.transports.clear();
+
+    OrchestratorLogger.log(`[HotRunner] Service shut down`);
+  }
+
+  /**
+   * Get the status of all registered runners.
+   */
+  getStatus(): HotRunnerStatus[] {
+    return this.registry.listRunners();
+  }
+
+  /**
+   * Get the underlying registry (for testing or advanced use).
+   */
+  getRegistry(): HotRunnerRegistry {
+    return this.registry;
+  }
+
+  /**
+   * Parse a space-separated custom parameters string into a key-value map.
+   * Handles `-key value` and `-key=value` formats.
+   */
+  private parseCustomParameters(raw: string): Record<string, string> {
+    const result: Record<string, string> = {};
+    const parts = raw.trim().split(/\s+/);
+
+    for (let i = 0; i < parts.length; i++) {
+      const part = parts[i];
+      if (part.startsWith('-')) {
+        const key = part.replace(/^-+/, '');
+        if (key.includes('=')) {
+          const [k, ...v] = key.split('=');
+          result[k] = v.join('=');
+        } else if (i + 1 < parts.length && !parts[i + 1].startsWith('-')) {
+          result[key] = parts[i + 1];
+          i++;
+        } else {
+          result[key] = 'true';
+        }
+      }
+    }
+
+    return result;
+  }
+}

src/model/orchestrator/services/hot-runner/hot-runner-types.ts

@@ -0,0 +1,54 @@
+export interface HotRunnerConfig {
+  enabled: boolean;
+  transport: 'websocket' | 'grpc' | 'named-pipe';
+  host: string;
+  port: number;
+  healthCheckInterval: number; // seconds
+  maxIdleTime: number; // seconds before recycling
+  maxJobsBeforeRecycle: number;
+  unityVersion?: string;
+  platform?: string;
+}
+
+export interface HotRunnerStatus {
+  id: string;
+  state: 'idle' | 'busy' | 'starting' | 'stopping' | 'unhealthy';
+  unityVersion: string;
+  platform: string;
+  currentJob?: string;
+  lastJobCompleted?: string;
+  uptime: number;
+  jobsCompleted: number;
+  lastHealthCheck: string;
+  memoryUsageMB: number;
+  libraryHash?: string;
+}
+
+export interface HotRunnerJobRequest {
+  jobId: string;
+  buildMethod?: string;
+  buildTarget: string;
+  buildPath?: string;
+  customParameters?: Record<string, string>;
+  timeout: number;
+  testMode?: 'editmode' | 'playmode';
+  testSuitePath?: string;
+}
+
+export interface HotRunnerJobResult {
+  jobId: string;
+  success: boolean;
+  exitCode: number;
+  duration: number;
+  output: string;
+  artifacts?: string[];
+  testResults?: string; // path to test result file
+}
+
+export interface HotRunnerTransport {
+  connect(config: HotRunnerConfig): Promise<void>;
+  disconnect(): Promise<void>;
+  sendJob(request: HotRunnerJobRequest): Promise<HotRunnerJobResult>;
+  getStatus(): Promise<HotRunnerStatus>;
+  healthCheck(): Promise<boolean>;
+}

src/model/orchestrator/services/hot-runner/hot-runner.test.ts

@@ -0,0 +1,740 @@
+import fs from 'node:fs';
+import { HotRunnerRegistry } from './hot-runner-registry';
+import { HotRunnerHealthMonitor } from './hot-runner-health-monitor';
+import { HotRunnerDispatcher } from './hot-runner-dispatcher';
+import { HotRunnerService } from './hot-runner-service';
+import {
+  HotRunnerConfig,
+  HotRunnerJobRequest,
+  HotRunnerJobResult,
+  HotRunnerStatus,
+  HotRunnerTransport,
+} from './hot-runner-types';
+
+// Mock dependencies
+jest.mock('node:fs');
+jest.mock('../core/orchestrator-logger');
+
+const mockFs = fs as jest.Mocked<typeof fs>;
+
+function createMockConfig(overrides?: Partial<HotRunnerConfig>): HotRunnerConfig {
+  return {
+    enabled: true,
+    transport: 'websocket',
+    host: 'localhost',
+    port: 9090,
+    healthCheckInterval: 30,
+    maxIdleTime: 3600,
+    maxJobsBeforeRecycle: 100,
+    unityVersion: '2022.3.0f1',
+    platform: 'StandaloneWindows64',
+    ...overrides,
+  };
+}
+
+function createMockTransport(overrides?: Partial<HotRunnerTransport>): HotRunnerTransport {
+  return {
+    // eslint-disable-next-line unicorn/no-useless-undefined
+    connect: jest.fn().mockResolvedValue(undefined),
+    // eslint-disable-next-line unicorn/no-useless-undefined
+    disconnect: jest.fn().mockResolvedValue(undefined),
+    sendJob: jest.fn().mockResolvedValue({
+      jobId: 'test-job',
+      success: true,
+      exitCode: 0,
+      duration: 5000,
+      output: 'Build succeeded',
+      artifacts: ['build/output.exe'],
+    } as HotRunnerJobResult),
+    getStatus: jest.fn().mockResolvedValue({
+      id: 'mock-runner',
+      state: 'idle',
+      unityVersion: '2022.3.0f1',
+      platform: 'StandaloneWindows64',
+      uptime: 3600,
+      jobsCompleted: 5,
+      lastHealthCheck: new Date().toISOString(),
+      memoryUsageMB: 1024,
+    } as HotRunnerStatus),
+    healthCheck: jest.fn().mockResolvedValue(true),
+    ...overrides,
+  };
+}
+
+function createMockJobRequest(overrides?: Partial<HotRunnerJobRequest>): HotRunnerJobRequest {
+  return {
+    jobId: 'job-001',
+    buildTarget: 'StandaloneWindows64',
+    timeout: 60000,
+    ...overrides,
+  };
+}
+
+// --- Registry Tests ---
+
+describe('HotRunnerRegistry', () => {
+  let registry: HotRunnerRegistry;
+
+  beforeEach(() => {
+    jest.clearAllMocks();
+    registry = new HotRunnerRegistry();
+  });
+
+  it('should register a runner and return an ID', () => {
+    const config = createMockConfig();
+    const id = registry.registerRunner(config);
+
+    expect(id).toMatch(/^hr-/);
+    expect(registry.size).toBe(1);
+  });
+
+  it('should retrieve a registered runner by ID', () => {
+    const config = createMockConfig();
+    const id = registry.registerRunner(config);
+    const runner = registry.getRunner(id);
+
+    expect(runner).toBeDefined();
+    expect(runner!.id).toBe(id);
+    expect(runner!.state).toBe('starting');
+    expect(runner!.unityVersion).toBe('2022.3.0f1');
+    expect(runner!.platform).toBe('StandaloneWindows64');
+  });
+
+  it('should return undefined for unknown runner ID', () => {
+    const runner = registry.getRunner('nonexistent');
+    expect(runner).toBeUndefined();
+  });
+
+  it('should unregister a runner', () => {
+    const id = registry.registerRunner(createMockConfig());
+    expect(registry.size).toBe(1);
+
+    registry.unregisterRunner(id);
+    expect(registry.size).toBe(0);
+    expect(registry.getRunner(id)).toBeUndefined();
+  });
+
+  it('should handle unregistering a nonexistent runner gracefully', () => {
+    registry.unregisterRunner('nonexistent');
+    expect(registry.size).toBe(0);
+  });
+
+  it('should list all runners without filter', () => {
+    registry.registerRunner(createMockConfig({ platform: 'StandaloneWindows64' }));
+    registry.registerRunner(createMockConfig({ platform: 'StandaloneLinux64' }));
+    registry.registerRunner(createMockConfig({ platform: 'StandaloneOSX' }));
+
+    const all = registry.listRunners();
+    expect(all).toHaveLength(3);
+  });
+
+  it('should filter runners by platform', () => {
+    registry.registerRunner(createMockConfig({ platform: 'StandaloneWindows64' }));
+    registry.registerRunner(createMockConfig({ platform: 'StandaloneLinux64' }));
+    registry.registerRunner(createMockConfig({ platform: 'StandaloneWindows64' }));
+
+    const windows = registry.listRunners({ platform: 'StandaloneWindows64' });
+    expect(windows).toHaveLength(2);
+
+    const linux = registry.listRunners({ platform: 'StandaloneLinux64' });
+    expect(linux).toHaveLength(1);
+  });
+
+  it('should filter runners by state', () => {
+    const id1 = registry.registerRunner(createMockConfig());
+    registry.registerRunner(createMockConfig());
+
+    registry.updateRunner(id1, { state: 'idle' });
+
+    // second runner remains in 'starting' state
+
+    const idle = registry.listRunners({ state: 'idle' });
+    expect(idle).toHaveLength(1);
+    expect(idle[0].id).toBe(id1);
+  });
+
+  it('should filter runners by Unity version', () => {
+    registry.registerRunner(createMockConfig({ unityVersion: '2022.3.0f1' }));
+    registry.registerRunner(createMockConfig({ unityVersion: '2023.1.0f1' }));
+    registry.registerRunner(createMockConfig({ unityVersion: '2022.3.0f1' }));
+
+    const v2022 = registry.listRunners({ unityVersion: '2022.3.0f1' });
+    expect(v2022).toHaveLength(2);
+  });
+
+  it('should find an available idle runner matching requirements', () => {
+    const id1 = registry.registerRunner(
+      createMockConfig({ unityVersion: '2022.3.0f1', platform: 'StandaloneWindows64' }),
+    );
+    registry.updateRunner(id1, { state: 'idle' });
+
+    const id2 = registry.registerRunner(
+      createMockConfig({ unityVersion: '2023.1.0f1', platform: 'StandaloneLinux64' }),
+    );
+    registry.updateRunner(id2, { state: 'idle' });
+
+    const found = registry.findAvailableRunner({
+      unityVersion: '2022.3.0f1',
+      platform: 'StandaloneWindows64',
+    });
+
+    expect(found).toBeDefined();
+    expect(found!.id).toBe(id1);
+  });
+
+  it('should return undefined when no runner matches requirements', () => {
+    const id = registry.registerRunner(
+      createMockConfig({ unityVersion: '2022.3.0f1', platform: 'StandaloneWindows64' }),
+    );
+    registry.updateRunner(id, { state: 'idle' });
+
+    const found = registry.findAvailableRunner({
+      unityVersion: '2023.1.0f1',
+      platform: 'StandaloneLinux64',
+    });
+
+    expect(found).toBeUndefined();
+  });
+
+  it('should update runner status fields', () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle', memoryUsageMB: 2048 });
+
+    const runner = registry.getRunner(id);
+    expect(runner!.state).toBe('idle');
+    expect(runner!.memoryUsageMB).toBe(2048);
+
+    // ID should not be overridden by the update
+    expect(runner!.id).toBe(id);
+  });
+
+  it('should persist and load registry from disk', () => {
+    const persistenceRegistry = new HotRunnerRegistry('/tmp/test');
+    mockFs.existsSync.mockReturnValue(true);
+    mockFs.writeFileSync.mockImplementation(() => {});
+    mockFs.mkdirSync.mockImplementation(() => '' as any);
+
+    const id = persistenceRegistry.registerRunner(createMockConfig());
+
+    // Verify writeFileSync was called for persistence
+    expect(mockFs.writeFileSync).toHaveBeenCalled();
+    const writtenData = JSON.parse((mockFs.writeFileSync as jest.Mock).mock.calls[0][1] as string);
+    expect(writtenData.runners).toBeDefined();
+    expect(writtenData.runners[id]).toBeDefined();
+  });
+
+  it('should load runners from disk on loadFromDisk', () => {
+    const persistenceRegistry = new HotRunnerRegistry('/tmp/test');
+    const storedData = {
+      runners: {
+        'hr-restored': {
+          id: 'hr-restored',
+          state: 'idle',
+          unityVersion: '2022.3.0f1',
+          platform: 'StandaloneWindows64',
+          uptime: 100,
+          jobsCompleted: 3,
+          lastHealthCheck: new Date().toISOString(),
+          memoryUsageMB: 512,
+        },
+      },
+      configs: {
+        'hr-restored': createMockConfig(),
+      },
+    };
+
+    mockFs.existsSync.mockReturnValue(true);
+    mockFs.readFileSync.mockReturnValue(JSON.stringify(storedData));
+
+    const count = persistenceRegistry.loadFromDisk();
+    expect(count).toBe(1);
+    expect(persistenceRegistry.getRunner('hr-restored')).toBeDefined();
+  });
+
+  it('should discard invalid runner entries when loading from disk', () => {
+    const persistenceRegistry = new HotRunnerRegistry('/tmp/test');
+    const storedData = {
+      runners: {
+        'hr-valid': {
+          id: 'hr-valid',
+          state: 'idle',
+          unityVersion: '2022.3.0f1',
+          platform: 'StandaloneWindows64',
+          uptime: 100,
+          jobsCompleted: 3,
+          lastHealthCheck: new Date().toISOString(),
+          memoryUsageMB: 512,
+        },
+        'hr-invalid': {
+          // Missing required fields like state, unityVersion
+          id: 'hr-invalid',
+        },
+        'hr-bad-state': {
+          id: 'hr-bad-state',
+          state: 'nonexistent-state',
+          unityVersion: '2022.3.0f1',
+          platform: 'StandaloneWindows64',
+          uptime: 0,
+          jobsCompleted: 0,
+          lastHealthCheck: new Date().toISOString(),
+          memoryUsageMB: 0,
+        },
+      },
+      configs: {
+        'hr-valid': createMockConfig(),
+        'hr-invalid': createMockConfig(),
+        'hr-bad-state': createMockConfig(),
+      },
+    };
+
+    mockFs.existsSync.mockReturnValue(true);
+    mockFs.readFileSync.mockReturnValue(JSON.stringify(storedData));
+
+    const count = persistenceRegistry.loadFromDisk();
+    expect(count).toBe(1);
+    expect(persistenceRegistry.getRunner('hr-valid')).toBeDefined();
+    expect(persistenceRegistry.getRunner('hr-invalid')).toBeUndefined();
+    expect(persistenceRegistry.getRunner('hr-bad-state')).toBeUndefined();
+  });
+
+  it('should handle corrupt JSON persistence file gracefully', () => {
+    const persistenceRegistry = new HotRunnerRegistry('/tmp/test');
+
+    mockFs.existsSync.mockReturnValue(true);
+    mockFs.readFileSync.mockReturnValue('{ invalid json !!!');
+
+    const count = persistenceRegistry.loadFromDisk();
+    expect(count).toBe(0);
+    expect(persistenceRegistry.size).toBe(0);
+  });
+
+  it('should handle persistence file with invalid top-level structure', () => {
+    const persistenceRegistry = new HotRunnerRegistry('/tmp/test');
+
+    mockFs.existsSync.mockReturnValue(true);
+    mockFs.readFileSync.mockReturnValue('"just a string"');
+
+    const count = persistenceRegistry.loadFromDisk();
+    expect(count).toBe(0);
+  });
+
+  it('should handle persistence file with null runners', () => {
+    const persistenceRegistry = new HotRunnerRegistry('/tmp/test');
+
+    mockFs.existsSync.mockReturnValue(true);
+    mockFs.readFileSync.mockReturnValue('{"runners": null, "configs": null}');
+
+    const count = persistenceRegistry.loadFromDisk();
+    expect(count).toBe(0);
+  });
+
+  it('should validate and repair invalid runners', () => {
+    const persistenceRegistry = new HotRunnerRegistry('/tmp/test');
+    mockFs.existsSync.mockReturnValue(true);
+    mockFs.writeFileSync.mockImplementation(() => {});
+    mockFs.mkdirSync.mockImplementation(() => '' as any);
+
+    // Register a valid runner first
+    const id = persistenceRegistry.registerRunner(createMockConfig());
+    persistenceRegistry.updateRunner(id, { state: 'idle' });
+
+    // Manually corrupt the runner's state by setting an invalid state
+    // (we access via the public API -- updateRunner with a cast)
+    persistenceRegistry.updateRunner(id, { state: 'invalid-state' as any });
+
+    const repaired = persistenceRegistry.validateAndRepair();
+    expect(repaired).toBe(1);
+
+    const runner = persistenceRegistry.getRunner(id);
+    expect(runner!.state).toBe('unhealthy');
+  });
+
+  it('should not discard configs for valid runners when loading from disk', () => {
+    const persistenceRegistry = new HotRunnerRegistry('/tmp/test');
+    const storedData = {
+      runners: {
+        'hr-valid': {
+          id: 'hr-valid',
+          state: 'idle',
+          unityVersion: '2022.3.0f1',
+          platform: 'StandaloneWindows64',
+          uptime: 100,
+          jobsCompleted: 3,
+          lastHealthCheck: new Date().toISOString(),
+          memoryUsageMB: 512,
+        },
+      },
+      configs: {
+        'hr-valid': createMockConfig(),
+      },
+    };
+
+    mockFs.existsSync.mockReturnValue(true);
+    mockFs.readFileSync.mockReturnValue(JSON.stringify(storedData));
+
+    persistenceRegistry.loadFromDisk();
+    expect(persistenceRegistry.getConfig('hr-valid')).toBeDefined();
+  });
+});
+
+// --- Health Monitor Tests ---
+
+describe('HotRunnerHealthMonitor', () => {
+  let monitor: HotRunnerHealthMonitor;
+  let registry: HotRunnerRegistry;
+  let transports: Map<string, HotRunnerTransport>;
+
+  beforeEach(() => {
+    jest.clearAllMocks();
+    jest.useFakeTimers();
+    monitor = new HotRunnerHealthMonitor();
+    registry = new HotRunnerRegistry();
+    transports = new Map();
+  });
+
+  afterEach(() => {
+    monitor.stopMonitoring();
+    jest.useRealTimers();
+  });
+
+  it('should start and stop monitoring', () => {
+    monitor.startMonitoring(registry, 30, transports);
+    expect(monitor.isMonitoring).toBe(true);
+
+    monitor.stopMonitoring();
+    expect(monitor.isMonitoring).toBe(false);
+  });
+
+  it('should report healthy when transport health check passes', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const transport = createMockTransport();
+    transports.set(id, transport);
+    monitor.startMonitoring(registry, 30, transports);
+
+    const healthy = await monitor.checkHealth(id);
+    expect(healthy).toBe(true);
+    expect(transport.healthCheck).toHaveBeenCalled();
+  });
+
+  it('should mark runner as unhealthy when health check fails', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const transport = createMockTransport({
+      healthCheck: jest.fn().mockResolvedValue(false),
+    });
+    transports.set(id, transport);
+    monitor.startMonitoring(registry, 30, transports);
+
+    const healthy = await monitor.checkHealth(id);
+    expect(healthy).toBe(false);
+
+    const runner = registry.getRunner(id);
+    expect(runner!.state).toBe('unhealthy');
+  });
+
+  it('should mark runner as unhealthy when health check throws', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const transport = createMockTransport({
+      healthCheck: jest.fn().mockRejectedValue(new Error('Connection refused')),
+    });
+    transports.set(id, transport);
+    monitor.startMonitoring(registry, 30, transports);
+
+    const healthy = await monitor.checkHealth(id);
+    expect(healthy).toBe(false);
+  });
+
+  it('should recycle unhealthy runner and remove from registry', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    const transport = createMockTransport();
+    transports.set(id, transport);
+    monitor.startMonitoring(registry, 30, transports);
+
+    await monitor.recycleUnhealthyRunner(id);
+
+    expect(registry.getRunner(id)).toBeUndefined();
+    expect(transport.disconnect).toHaveBeenCalled();
+    expect(transports.has(id)).toBe(false);
+  });
+
+  it('should recycle idle runner when max idle time exceeded', async () => {
+    const id = registry.registerRunner(createMockConfig({ maxIdleTime: 60 }));
+
+    // Set lastHealthCheck to 120 seconds ago
+    const oldDate = new Date(Date.now() - 120 * 1000).toISOString();
+    registry.updateRunner(id, { state: 'idle', lastHealthCheck: oldDate });
+
+    const transport = createMockTransport();
+    transports.set(id, transport);
+    monitor.startMonitoring(registry, 30, transports);
+
+    await monitor.recycleIdleRunner(id, 60);
+
+    expect(registry.getRunner(id)).toBeUndefined();
+  });
+
+  it('should not recycle idle runner when within max idle time', async () => {
+    const id = registry.registerRunner(createMockConfig({ maxIdleTime: 3600 }));
+    registry.updateRunner(id, {
+      state: 'idle',
+      lastHealthCheck: new Date().toISOString(),
+    });
+
+    const transport = createMockTransport();
+    transports.set(id, transport);
+    monitor.startMonitoring(registry, 30, transports);
+
+    await monitor.recycleIdleRunner(id, 3600);
+
+    // Runner should still exist
+    expect(registry.getRunner(id)).toBeDefined();
+  });
+
+  it('should return false when no transport exists for runner', async () => {
+    const id = registry.registerRunner(createMockConfig());
+
+    // Do not set any transport for this runner
+    monitor.startMonitoring(registry, 30, transports);
+
+    const healthy = await monitor.checkHealth(id);
+    expect(healthy).toBe(false);
+  });
+});
+
+// --- Dispatcher Tests ---
+
+describe('HotRunnerDispatcher', () => {
+  let registry: HotRunnerRegistry;
+  let transports: Map<string, HotRunnerTransport>;
+  let dispatcher: HotRunnerDispatcher;
+
+  beforeEach(() => {
+    jest.clearAllMocks();
+    registry = new HotRunnerRegistry();
+    transports = new Map();
+    dispatcher = new HotRunnerDispatcher(transports);
+  });
+
+  it('should dispatch a job to an available runner', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const transport = createMockTransport();
+    transports.set(id, transport);
+
+    const request = createMockJobRequest();
+    const result = await dispatcher.dispatchJob(request, registry, '2022.3.0f1');
+
+    expect(result.success).toBe(true);
+    expect(result.exitCode).toBe(0);
+    expect(transport.sendJob).toHaveBeenCalledWith(request);
+  });
+
+  it('should mark runner as busy during job execution', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const statesDuringJob: string[] = [];
+    const transport = createMockTransport({
+      sendJob: jest.fn().mockImplementation(async () => {
+        const runner = registry.getRunner(id);
+        if (runner) statesDuringJob.push(runner.state);
+
+        return {
+          jobId: 'job-001',
+          success: true,
+          exitCode: 0,
+          duration: 1000,
+          output: 'ok',
+        };
+      }),
+    });
+    transports.set(id, transport);
+
+    await dispatcher.dispatchJob(createMockJobRequest(), registry, '2022.3.0f1');
+
+    expect(statesDuringJob).toContain('busy');
+
+    // After completion, should be idle again
+    const runner = registry.getRunner(id);
+    expect(runner!.state).toBe('idle');
+  });
+
+  it('should increment jobsCompleted after successful dispatch', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle', jobsCompleted: 5 });
+
+    const transport = createMockTransport();
+    transports.set(id, transport);
+
+    await dispatcher.dispatchJob(createMockJobRequest(), registry, '2022.3.0f1');
+
+    const runner = registry.getRunner(id);
+    expect(runner!.jobsCompleted).toBe(6);
+  });
+
+  it('should throw when no runner is available and wait times out', async () => {
+    // No runners registered at all
+    const request = createMockJobRequest({ timeout: 100 });
+
+    await expect(dispatcher.dispatchJob(request, registry, '2022.3.0f1')).rejects.toThrow(/Timed out waiting/);
+  });
+
+  it('should throw when runner has no transport', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    // No transport set for this runner
+
+    const request = createMockJobRequest();
+
+    await expect(dispatcher.dispatchJob(request, registry, '2022.3.0f1')).rejects.toThrow(/No transport available/);
+  });
+
+  it('should handle job failure and return runner to idle', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const transport = createMockTransport({
+      sendJob: jest.fn().mockRejectedValue(new Error('Unity crashed')),
+    });
+    transports.set(id, transport);
+
+    await expect(dispatcher.dispatchJob(createMockJobRequest(), registry, '2022.3.0f1')).rejects.toThrow(
+      'Unity crashed',
+    );
+
+    // Runner should be back to idle despite failure
+    const runner = registry.getRunner(id);
+    expect(runner!.state).toBe('idle');
+  });
+
+  it('should handle job timeout', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const transport = createMockTransport({
+      sendJob: jest.fn().mockImplementation(
+        () => new Promise((resolve) => setTimeout(resolve, 60000)), // never resolves within timeout
+      ),
+    });
+    transports.set(id, transport);
+
+    const request = createMockJobRequest({ timeout: 50 });
+
+    await expect(dispatcher.dispatchJob(request, registry, '2022.3.0f1')).rejects.toThrow(/timed out/);
+  });
+
+  it('should disconnect transport on job timeout', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const transport = createMockTransport({
+      sendJob: jest.fn().mockImplementation(
+        () => new Promise((resolve) => setTimeout(resolve, 60000)), // never resolves within timeout
+      ),
+    });
+    transports.set(id, transport);
+
+    const request = createMockJobRequest({ timeout: 50 });
+
+    await expect(dispatcher.dispatchJob(request, registry, '2022.3.0f1')).rejects.toThrow(/timed out/);
+
+    // Transport should have been disconnected to clean up orphaned connection
+    expect(transport.disconnect).toHaveBeenCalled();
+  });
+
+  it('should call output callback with job output', async () => {
+    const id = registry.registerRunner(createMockConfig());
+    registry.updateRunner(id, { state: 'idle' });
+
+    const transport = createMockTransport();
+    transports.set(id, transport);
+
+    const outputCallback = jest.fn();
+    await dispatcher.dispatchJob(createMockJobRequest(), registry, '2022.3.0f1', outputCallback);
+
+    expect(outputCallback).toHaveBeenCalledWith('Build succeeded');
+  });
+
+  it('should wait for runner to become available', async () => {
+    const id = registry.registerRunner(createMockConfig());
+
+    // Runner starts in 'starting' state, not idle
+
+    const transport = createMockTransport();
+    transports.set(id, transport);
+
+    // Simulate runner becoming idle after a short delay
+    setTimeout(() => {
+      registry.updateRunner(id, { state: 'idle' });
+    }, 50);
+
+    const request = createMockJobRequest({ timeout: 5000 });
+    const result = await dispatcher.dispatchJob(request, registry, '2022.3.0f1');
+
+    expect(result.success).toBe(true);
+  });
+});
+
+// --- Service Integration Tests ---
+
+describe('HotRunnerService', () => {
+  let service: HotRunnerService;
+
+  beforeEach(() => {
+    jest.clearAllMocks();
+    mockFs.existsSync.mockReturnValue(false);
+    service = new HotRunnerService();
+  });
+
+  afterEach(async () => {
+    await service.shutdown();
+  });
+
+  it('should initialize and shut down cleanly', async () => {
+    const config = createMockConfig();
+    await service.initialize(config);
+
+    const status = service.getStatus();
+    expect(status).toEqual([]);
+
+    await service.shutdown();
+  });
+
+  it('should register a runner with transport', async () => {
+    await service.initialize(createMockConfig());
+
+    const transport = createMockTransport();
+    const id = service.registerRunner(createMockConfig(), transport);
+
+    expect(id).toMatch(/^hr-/);
+    expect(service.getStatus()).toHaveLength(1);
+  });
+
+  it('should disconnect all transports on shutdown', async () => {
+    await service.initialize(createMockConfig());
+
+    const transport1 = createMockTransport();
+    const transport2 = createMockTransport();
+    service.registerRunner(createMockConfig(), transport1);
+    service.registerRunner(createMockConfig(), transport2);
+
+    await service.shutdown();
+
+    expect(transport1.disconnect).toHaveBeenCalled();
+    expect(transport2.disconnect).toHaveBeenCalled();
+  });
+
+  it('should expose the underlying registry', async () => {
+    await service.initialize(createMockConfig());
+    const registry = service.getRegistry();
+
+    expect(registry).toBeInstanceOf(HotRunnerRegistry);
+  });
+});

src/model/orchestrator/services/hot-runner/index.ts

@@ -0,0 +1,11 @@
+export { HotRunnerService } from './hot-runner-service';
+export { HotRunnerRegistry } from './hot-runner-registry';
+export { HotRunnerHealthMonitor } from './hot-runner-health-monitor';
+export { HotRunnerDispatcher } from './hot-runner-dispatcher';
+export type {
+  HotRunnerConfig,
+  HotRunnerStatus,
+  HotRunnerJobRequest,
+  HotRunnerJobResult,
+  HotRunnerTransport,
+} from './hot-runner-types';