feat: Auto-retry failed tasks with exponential backoff

- Hard failures now auto-retry up to 3 times
- Exponential backoff: 5, 10, 20 minutes
- Only permanently fails after max retries exceeded
- Soft failures still requeue immediately

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

backend/src/tasks/task-service.ts

@@ -301,12 +301,20 @@ class TaskService {
   /**
    * Mark a task as failed
    *
-   * Soft failures (timeouts, connection issues): Requeue back to pending for later pickup
-   * Hard failures (business logic errors): Mark as failed permanently
+   * Soft failures (timeouts, connection issues): Requeue immediately
+   * Hard failures: Auto-retry up to MAX_RETRIES with exponential backoff
    */
   async failTask(taskId: number, errorMessage: string): Promise<boolean> {
+    const MAX_RETRIES = 3;
     const isSoft = this.isSoftFailure(errorMessage);
 
+    // Get current retry count
+    const { rows } = await pool.query(
+      `SELECT retry_count FROM worker_tasks WHERE id = $1`,
+      [taskId]
+    );
+    const retryCount = rows[0]?.retry_count ?? 0;
+
     if (isSoft) {
       // Soft failure: put back in queue immediately for another worker
       await pool.query(
@@ -325,16 +333,36 @@ class TaskService {
       return true;
     }
 
-    // Hard failure: mark as permanently failed
+    // Hard failure: auto-retry with exponential backoff if under max retries
+    if (retryCount < MAX_RETRIES) {
+      const delayMinutes = Math.pow(2, retryCount) * 5; // 5, 10, 20 minutes
+      await pool.query(
+        `UPDATE worker_tasks
+         SET status = 'pending',
+             worker_id = NULL,
+             claimed_at = NULL,
+             started_at = NULL,
+             error_message = $2,
+             retry_count = retry_count + 1,
+             scheduled_for = NOW() + INTERVAL '${delayMinutes} minutes',
+             updated_at = NOW()
+         WHERE id = $1`,
+        [taskId, `Retry ${retryCount + 1}/${MAX_RETRIES}: ${errorMessage}`]
+      );
+      console.log(`[TaskService] Task ${taskId} scheduled for retry ${retryCount + 1}/${MAX_RETRIES} in ${delayMinutes} minutes`);
+      return true;
+    }
+
+    // Max retries exceeded: mark as permanently failed
     await pool.query(
       `UPDATE worker_tasks
        SET status = 'failed',
            completed_at = NOW(),
            error_message = $2
        WHERE id = $1`,
-      [taskId, `Hard failure: ${errorMessage}`]
+      [taskId, `Failed after ${MAX_RETRIES} retries: ${errorMessage}`]
     );
-    console.log(`[TaskService] Task ${taskId} hard failed: ${errorMessage}`);
+    console.log(`[TaskService] Task ${taskId} permanently failed after ${MAX_RETRIES} retries`);
     return false;
   }