评测执行的测试不依赖同进程里的数据源个数、等待都有时限：别的测试建的数据源超过评测拍目录快照的上限时，作答运行在排队前就失败，原来的等待循环永远等不到、整次测试卡死（上传源测试和评测测试分到同一个进程时复现）；这里的评测不测这个上限，测试里把上限放大，等待改成 30 秒等不到就失败并说清在等什么

backend/tests/test_eval_runner.py

@@ -33,7 +33,12 @@ async def _clear_ledger() -> None:
 
 
 @pytest.fixture(autouse=True)
-async def engine_up():
+async def engine_up(monkeypatch):
+    # 同一个测试进程里别的测试建的数据源可能超过评测拍目录快照的上限（app/eval/snapshots.MAX_SOURCES），评测会按
+    # 设计报「数据源超过 20 个」——这里的评测不测这个上限，分到哪个进程都要照常跑
+    from app.eval import snapshots
+
+    monkeypatch.setattr(snapshots, "MAX_SOURCES", 10_000)
     await run_manager.setup()
     await _clear_ledger()
     yield
@@ -49,6 +54,16 @@ async def client():
         yield c
 
 
+async def _until(cond, what: str, *, timeout: float = 30.0) -> None:
+    """等到 cond() 为真；等不到就让测试失败、说清在等什么，不让整次测试卡死（作答运行在排队之前就失败时，排队永远等不到）。"""
+    loop = asyncio.get_running_loop()
+    deadline = loop.time() + timeout
+    while not await cond():
+        if loop.time() > deadline:
+            raise AssertionError(f"等了 {timeout:g} 秒还没等到：{what}")
+        await asyncio.sleep(0.02)
+
+
 def ask_cases(n: int) -> list[dict]:
     return [{"question": f"第 {i} 个含糊的问题", "expectations": [{"kind": "ask"}]} for i in range(1, n + 1)]
 
@@ -327,12 +342,13 @@ async def test_time_queued_for_a_run_slot_does_not_count_against_the_case_limit(
     monkeypatch.setattr(runner.EvalExecution, "_prepare", prepare)
     held = asyncio.create_task(blocker())
     try:
-        while not (await slots.snapshot())["classes"]["system"]["running"]:
+        async def blocking() -> bool:
-            await asyncio.sleep(0.02)
+            return bool((await slots.snapshot())["classes"]["system"]["running"])
+
+        await _until(blocking, "占名额的那一段拿到名额")
         suite_id = await es.make_suite("none", cases=ask_cases(1))
         [run] = await start(suite_id)
-        while not await slots.waiting("eval-answer-in-queue"):
+        await _until(lambda: slots.waiting("eval-answer-in-queue"), "作答运行排队等名额")
-            await asyncio.sleep(0.02)
         await asyncio.sleep(2.0)
         gate.set()
         await es.wait_eval(run.id)