fix: predictions row carries arm run_id under shared gate_log; drain evolve gather on failure (algo #6)

app/harness/inference.py

@@ -409,7 +409,11 @@ async def _run_single_question(
     返回:
         预测结果字典（含 video_id, question_id, prediction, answer 等）。
     """
+    # run_id 必须显式入 record:HarnessLog.insert 缺省用**实例** run_id 填充,
+    # 连续并发 gate 共享单一 gate_log(实例 run_id 为 step 级)时,各臂行必须
+    # 落自己的臂 run_id,否则 validate 回读 _load_run_rows(臂 run_id) 为空。
     record: dict[str, Any] = {
+        "run_id": run_id,
         "video_id": qa.video_id,
         "question_id": qa.question_id,
         "task_type": qa.task_type,

app/harness/runner.py

@@ -1370,13 +1370,17 @@ class Runner:
                 rejected=state.rejected_buffer.get(task_type, []),
             )
 
-        records = dict(
-            zip(
-                active_types,
-                await asyncio.gather(*[_evolve_one(t) for t in active_types]),
-                strict=True,
-            )
-        )
+        # 首异常先取消其余进化任务并排水再向上传播(与 validate_skills_concurrent
+        # 同款语义):避免失败后残留 in-flight LLM 任务与 pending task 警告。
+        tasks = [asyncio.ensure_future(_evolve_one(t)) for t in active_types]
+        try:
+            evolved = await asyncio.gather(*tasks)
+        except BaseException:
+            for task in tasks:
+                task.cancel()
+            await asyncio.gather(*tasks, return_exceptions=True)
+            raise
+        records = dict(zip(active_types, evolved, strict=True))
 
         # 无真实改动的题型照旧写 skipped 后出队
         gated: dict[str, EvolutionRecord] = {}