diff --git a/task_service.py b/task_service.py
index 69b4f98..de50d3d 100644
--- a/task_service.py
+++ b/task_service.py
@@ -522,6 +522,36 @@ def resume_task(public_id: str) -> dict[str, Any]:
     try:
         resumed = resume_collection(str(crawler_task_id))
     except CrawlerError as exc:
+        err_msg = str(exc)
+        # 常规/轻度任务（overview 类型）的爬虫任务不支持断点恢复，
+        # 回退到从初始化阶段重新执行，避免前端返回 503 错误。
+        if any(kw in err_msg for kw in ("不支持断点恢复", "sources 创建的任务")):
+            with connection() as conn:
+                row = conn.execute(
+                    "SELECT id FROM research_tasks WHERE public_task_id=?", (public_id,)
+                ).fetchone()
+                task_pk = row["id"] if row else None
+                now = now_iso()
+                conn.execute(
+                    """UPDATE research_tasks
+                       SET status='waiting',execution_stage='queued',progress_percent=0,
+                           cancel_requested_at=NULL,completed_at=NULL,error_code=NULL,
+                           error_message=NULL,result_json=NULL,updated_at=?,version=version+1
+                       WHERE public_task_id=?""",
+                    (now, public_id),
+                )
+                if task_pk:
+                    _event(
+                        conn, task_pk, "retried", "queued", 0, "user",
+                        payload={"resume_mode": "restart_initialization", "reason": "crawler_task_type_not_support_resume"},
+                    )
+            threading.Thread(
+                target=_execute_task,
+                args=(public_id,),
+                daemon=True,
+                name=f"holy-crab-retry-{public_id}",
+            ).start()
+            return get_task(public_id)
         raise TaskError(
             "COLLECTOR_UNAVAILABLE", f"爬虫任务恢复失败：{exc}", 503
         ) from exc