From da799620baa0fcf48ed784696e8427555a126596 Mon Sep 17 00:00:00 2001
From: summerview1997 <thomas@summerview.com.tw>
Date: Tue, 16 Jun 2026 09:20:57 +0800
Subject: [PATCH 1/2] Fail fast on systemic SearchQA rollout failures

---
 skillopt/envs/searchqa/rollout.py | 21 +++++++++++++++++----
 1 file changed, 17 insertions(+), 4 deletions(-)

diff --git a/skillopt/envs/searchqa/rollout.py b/skillopt/envs/searchqa/rollout.py
index ab7215d..83165b5 100644
--- a/skillopt/envs/searchqa/rollout.py
+++ b/skillopt/envs/searchqa/rollout.py
@@ -13,20 +13,31 @@ from __future__ import annotations
 import json
 import os
 import time
-import traceback
+from collections import Counter
 from concurrent.futures import FIRST_COMPLETED, ThreadPoolExecutor, wait
 
-from skillopt.model import chat_target, get_target_backend, is_target_exec_backend
+from skillopt.envs.searchqa.evaluator import evaluate
+from skillopt.model import chat_target, is_target_exec_backend
 from skillopt.model.codex_harness import prepare_workspace, render_skill_md, run_target_exec
 from skillopt.prompts import load_prompt
-from skillopt.envs.searchqa.evaluator import evaluate
-
 
 # ── Prompt templates ─────────────────────────────────────────────────────────
 
 _MAX_CONTEXT_CHARS = 6000
 
 
+def _raise_on_systemic_failure(results: list[dict]) -> None:
+    """Abort when all rollout rows failed before any agent response."""
+    if not results or not all(row.get("agent_ok") is False for row in results):
+        return
+    reasons = Counter(str(row.get("fail_reason") or "unknown error") for row in results)
+    common_reason, count = reasons.most_common(1)[0]
+    raise RuntimeError(
+        f"SearchQA rollout failed for all {len(results)} items before an agent "
+        f"response ({count}x): {common_reason}"
+    )
+
+
 def _truncate_context(context: str, max_chars: int = _MAX_CONTEXT_CHARS) -> str:
     """Truncate context at [DOC] boundaries to stay within budget."""
     if len(context) <= max_chars:
@@ -379,6 +390,7 @@ def run_batch(
 
     pending = [it for it in items if str(it["id"]) not in done_ids]
     if not pending:
+        _raise_on_systemic_failure(existing)
         return existing
 
     total = len(existing) + len(pending)
@@ -478,4 +490,5 @@ def run_batch(
         finally:
             ex.shutdown(wait=False, cancel_futures=True)
 
+    _raise_on_systemic_failure(results)
     return results

From 923becb00f549c4e28a933e7f726820d2cdfd530 Mon Sep 17 00:00:00 2001
From: summerview1997 <thomas@summerview.com.tw>
Date: Tue, 16 Jun 2026 09:21:08 +0800
Subject: [PATCH 2/2] Add SearchQA rollout fail-fast tests

---
 tests/test_searchqa_rollout_failfast.py | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
 create mode 100644 tests/test_searchqa_rollout_failfast.py

diff --git a/tests/test_searchqa_rollout_failfast.py b/tests/test_searchqa_rollout_failfast.py
new file mode 100644
index 0000000..ef2ef71
--- /dev/null
+++ b/tests/test_searchqa_rollout_failfast.py
@@ -0,0 +1,25 @@
+import json
+
+import pytest
+
+from skillopt.envs.searchqa.rollout import run_batch
+
+
+def test_cached_systemic_rollout_failure_aborts(tmp_path):
+    (tmp_path / "results.jsonl").write_text(
+        "\n".join([
+            json.dumps({"id": "1", "agent_ok": False, "fail_reason": "endpoint missing"}),
+            json.dumps({"id": "2", "agent_ok": False, "fail_reason": "endpoint missing"}),
+        ]),
+        encoding="utf-8",
+    )
+
+    with pytest.raises(RuntimeError, match="endpoint missing"):
+        run_batch([{"id": "1"}, {"id": "2"}], str(tmp_path), "skill")
+
+
+def test_cached_answered_wrong_rollout_does_not_abort(tmp_path):
+    result = {"id": "1", "agent_ok": True, "hard": 0, "fail_reason": "wrong answer"}
+    (tmp_path / "results.jsonl").write_text(json.dumps(result), encoding="utf-8")
+
+    assert run_batch([{"id": "1"}], str(tmp_path), "skill") == [result]