Fix dashboard startup blocking on recovery

packages/engine/src/executor.ts

@@ -317,6 +317,8 @@ export interface TaskExecutorOptions {
 export class TaskExecutor {
   private activeWorktrees = new Map<string, string>();
   private executing = new Set<string>();
+  /** Completed orphan recovery tasks currently running during startup. */
+  private recoveringCompleted = new Set<string>();
   /** Active agent sessions per task, used to terminate on pause and inject steering. */
   private activeSessions = new Map<string, {
     session: AgentSession;
@@ -381,7 +383,7 @@ export class TaskExecutor {
 
   /** Returns the set of task IDs currently being executed. */
   getExecutingTaskIds(): Set<string> {
-    return new Set(this.executing);
+    return new Set([...this.executing, ...this.recoveringCompleted]);
   }
 
   /**
@@ -748,8 +750,19 @@ export class TaskExecutor {
       // Fast-path: if the task already completed its work (all steps done),
       // move it directly to in-review instead of re-executing from scratch.
       if (this.isTaskWorkComplete(task)) {
+        if (this.recoveringCompleted.has(task.id)) {
+          executorLog.log(`${task.id} completed-task recovery already running - skipping duplicate startup recovery`);
+          continue;
+        }
         executorLog.log(`${task.id} is already complete — fast-pathing to in-review`);
-        await this.recoverCompletedTask(task);
+        this.recoveringCompleted.add(task.id);
+        void this.recoverCompletedTask(task)
+          .catch((err) =>
+            executorLog.error(`Failed to recover completed orphan ${task.id}:`, err),
+          )
+          .finally(() => {
+            this.recoveringCompleted.delete(task.id);
+          });
         continue;
       }
 

packages/engine/src/project-engine.ts

@@ -17,6 +17,7 @@ import { CronRunner, createAiPromptExecutor } from "./cron-runner.js";
 import { aiMergeTask } from "./merger.js";
 import { PRIORITY_MERGE } from "./concurrency.js";
 import { runtimeLog } from "./logger.js";
+import type { HeartbeatMonitor, HeartbeatTriggerScheduler } from "./agent-heartbeat.js";
 
 /**
  * Callback for processing pull-request merge strategy.
@@ -258,6 +259,21 @@ export class ProjectEngine {
     return this.automationStore;
   }
 
+  /** Get the project's working directory. */
+  getWorkingDirectory(): string {
+    return this.config.workingDirectory;
+  }
+
+  /** Get the HeartbeatMonitor from the underlying runtime, if initialized. */
+  getHeartbeatMonitor(): HeartbeatMonitor | undefined {
+    return this.runtime.getHeartbeatMonitor();
+  }
+
+  /** Get the HeartbeatTriggerScheduler from the underlying runtime, if initialized. */
+  getHeartbeatTriggerScheduler(): HeartbeatTriggerScheduler | undefined {
+    return this.runtime.getTriggerScheduler();
+  }
+
   /**
    * Enqueue a task ID for auto-merge if it is not already queued or active.
    * Exposed publicly so callers can integrate the engine's merge queue with

packages/engine/src/restart.integration.test.ts

@@ -418,14 +418,34 @@ describe("In-progress task resume after restart", () => {
     await executor.resumeOrphaned();
 
     expect(executeSpy).not.toHaveBeenCalled();
-    expect(store.updateTask).toHaveBeenCalledWith("FN-963", {
-      modifiedFiles: ["packages/dashboard/app/components/SettingsModal.tsx"],
+    await vi.waitFor(() => {
+      expect(store.updateTask).toHaveBeenCalledWith("FN-963", {
+        modifiedFiles: ["packages/dashboard/app/components/SettingsModal.tsx"],
+      });
+      expect(store.moveTask).toHaveBeenCalledWith("FN-963", "in-review");
+      expect(store.logEntry).toHaveBeenCalledWith(
+        "FN-963",
+        "Auto-recovered: task work was complete but stuck in in-progress — moved to in-review",
+      );
     });
-    expect(store.moveTask).toHaveBeenCalledWith("FN-963", "in-review");
-    expect(store.logEntry).toHaveBeenCalledWith(
-      "FN-963",
-      "Auto-recovered: task work was complete but stuck in in-progress — moved to in-review",
+  });
+
+  it("resumeOrphaned() does not block startup on completed-task recovery", async () => {
+    const store = createMockStore();
+    const completedTask = makeTask("FN-964", "in-progress", {
+      worktree: "/tmp/wt/FN-964",
+      baseCommitSha: "base123",
+      steps: makeSteps("done", "done", "skipped"),
+    });
+    store.listTasks.mockResolvedValue([completedTask]);
+
+    const executor = new TaskExecutor(store, "/tmp/test");
+    const recoverSpy = vi.spyOn(executor, "recoverCompletedTask").mockImplementation(
+      () => new Promise(() => {}),
     );
+
+    await expect(executor.resumeOrphaned()).resolves.toBeUndefined();
+    expect(recoverSpy).toHaveBeenCalledWith(completedTask);
   });
 
   it("resumeOrphaned() leaves no-progress no-task_done failures for self-healing", async () => {

packages/engine/src/runtimes/in-process-runtime.ts

@@ -466,7 +466,9 @@ export class InProcessRuntime
       // they no longer have a tracked session/worktree, so the stuck detector
       // cannot recover them. Delegate the startup recovery pass to
       // SelfHealingManager so the policy lives in one place.
-      await this.selfHealingManager.runStartupRecovery();
+      void this.selfHealingManager.runStartupRecovery().catch((err) => {
+        runtimeLog.error("Self-healing startup recovery failed:", err);
+      });
 
       // 11. Start scheduler and triage processor
       this.scheduler.start();