fix(agent-office/lotto): deep CuratorError fallthrough + urgent 발송 재시도

결함1: deep signal-check에서 curate_weekly가 CuratorError면 전체 check가 abort돼 sim/drift 시그널이 미평가되던 문제 → try/except로 confidence만 포기하고 sim/drift는 계속(curate_result=None fallthrough).
결함2: send_urgent_signal 실패가 outer except로 빠져 task 실패+미마킹이던 문제 → _send_urgent_with_retry(3회/60s) 추출, 최종 실패해도 raise 안 함(시그널 평가·태스크 보존), 성공 시에만 mark_signal_notified. TDD 3 신규 테스트.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

agent-office/app/agents/lotto.py

@@ -2,6 +2,10 @@ from .base import BaseAgent
 from ..db import create_task, update_task_status, add_log
 from ..curator.pipeline import curate_weekly, CuratorError
 
+# urgent 텔레그램 발송 재시도 (전송 실패가 시그널 평가/태스크를 중단시키지 않도록)
+URGENT_SEND_MAX_ATTEMPTS = 3
+URGENT_SEND_RETRY_SEC = 60
+
 
 class LottoAgent(BaseAgent):
     agent_id = "lotto"
@@ -54,10 +58,16 @@ class LottoAgent(BaseAgent):
 
             if source == "deep":
                 from ..curator.pipeline import curate_weekly
+                try:
                     cw = await curate_weekly(source="signal_deep")
                     curate_result = {"confidence": cw.get("confidence")}
                     if cw.get("draw_no"):
                         current_draw_no = cw.get("draw_no")
+                except CuratorError as e:
+                    # 큐레이션 실패는 confidence 시그널만 포기 — sim/drift 평가는 계속(fallthrough)
+                    add_log("lotto", f"deep curate_weekly 실패 → sim/drift만 평가: {e}",
+                            level="warning", task_id=task_id)
+                    curate_result = None
 
             outcome = await run_signal_check(
                 source=source,
@@ -88,11 +98,7 @@ class LottoAgent(BaseAgent):
                             "triggered_at": datetime.now(timezone.utc).isoformat(),
                             "results": outcome["results"],
                         }
-                        await send_urgent_signal(event)
+                        await self._send_urgent_with_retry(event, outcome["results"], task_id)
-                        for r in outcome["results"]:
-                            if r["fire_level"] in ("normal", "urgent"):
-                                mark_signal_notified(r["signal_id"])
-                        add_log("lotto", f"urgent 텔레그램 발송 ({len(outcome['results'])}개 시그널)", task_id=task_id)
 
             fired_metrics = [
                 r["metric"] for r in outcome["results"]
@@ -111,6 +117,31 @@ class LottoAgent(BaseAgent):
             add_log("lotto", f"signal_check 예외: {e}", level="error", task_id=task_id)
             return {"ok": False, "message": f"{type(e).__name__}: {e}"}
 
+    async def _send_urgent_with_retry(self, event: dict, results: list, task_id: str) -> bool:
+        """urgent 텔레그램 발송 + 실패 시 재시도. 최종 실패해도 raise하지 않음(시그널 평가·태스크 보존).
+        성공 시 fired 시그널을 notified로 마킹. 반환: 발송 성공 여부."""
+        import asyncio
+        from ..db import add_log, mark_signal_notified
+        from ..notifiers.telegram_lotto import send_urgent_signal
+        for attempt in range(1, URGENT_SEND_MAX_ATTEMPTS + 1):
+            try:
+                await send_urgent_signal(event)
+                for r in results:
+                    if r["fire_level"] in ("normal", "urgent"):
+                        mark_signal_notified(r["signal_id"])
+                add_log("lotto", f"urgent 텔레그램 발송 ({len(results)}개 시그널, attempt {attempt})", task_id=task_id)
+                return True
+            except Exception as e:
+                if attempt < URGENT_SEND_MAX_ATTEMPTS:
+                    add_log("lotto", f"urgent 발송 실패(attempt {attempt}) → {URGENT_SEND_RETRY_SEC}s 후 재시도: {e}",
+                            level="warning", task_id=task_id)
+                    await asyncio.sleep(URGENT_SEND_RETRY_SEC)
+                else:
+                    add_log("lotto", f"urgent 발송 {URGENT_SEND_MAX_ATTEMPTS}회 실패 — 미발송: {e}",
+                            level="error", task_id=task_id)
+                    return False
+        return False
+
     async def run_daily_digest(self) -> dict:
         """일일 요약 — 지난 24h normal/urgent 발화 텔레그램 1통. task_id wrap."""
         from ..db import (

agent-office/tests/test_lotto_task_wrap.py

@@ -96,6 +96,81 @@ async def test_run_signal_check_failure_marks_task_failed(monkeypatch):
     assert "boom" in tasks[0]["result_data"]["error"]
 
 
+@pytest.mark.asyncio
+async def test_deep_curate_error_still_evaluates_signals(monkeypatch):
+    """deep: curate_weekly가 CuratorError여도 sim/drift 시그널 평가는 계속(fallthrough)."""
+    from app.agents.lotto import LottoAgent
+    from app.curator import signal_runner, pipeline
+    from app import service_proxy
+    from app.notifiers import telegram_lotto
+
+    async def boom_curate(**kwargs):
+        raise pipeline.CuratorError("curation 실패")
+    monkeypatch.setattr(pipeline, "curate_weekly", boom_curate)
+
+    called = {"signal": False, "curate_result": "UNSET"}
+    async def fake_signal(**kwargs):
+        called["signal"] = True
+        called["curate_result"] = kwargs.get("curate_result")
+        return {"overall_fire": "normal", "results": [
+            {"signal_id": 1, "metric": "sim_signal", "value": 0.6, "z_score": 1.7,
+             "fire_level": "normal", "baseline_mu": 0.5, "baseline_sigma": 0.05, "payload": {}}]}
+    monkeypatch.setattr(signal_runner, "run_signal_check", fake_signal)
+
+    async def fake_latest():
+        return 1226
+    monkeypatch.setattr(service_proxy, "lotto_latest_draw", fake_latest)
+    async def fake_send(_e):
+        pass
+    monkeypatch.setattr(telegram_lotto, "send_urgent_signal", fake_send)
+
+    agent = LottoAgent()
+    result = await agent.run_signal_check(source="deep")
+    assert result["ok"] is True            # CuratorError로 중단되지 않음
+    assert called["signal"] is True        # sim/drift 평가 계속됨
+    assert called["curate_result"] is None  # confidence는 None으로 fallthrough
+
+
+@pytest.mark.asyncio
+async def test_urgent_send_retries_then_succeeds(monkeypatch):
+    """urgent 발송이 실패하면 재시도하고, 성공하면 True."""
+    from app.agents.lotto import LottoAgent
+    from app.notifiers import telegram_lotto
+    import app.agents.lotto as lotto_mod
+    monkeypatch.setattr(lotto_mod, "URGENT_SEND_RETRY_SEC", 0)  # 실대기 제거
+
+    attempts = {"n": 0}
+    async def flaky_send(_event):
+        attempts["n"] += 1
+        if attempts["n"] < 3:
+            raise RuntimeError("telegram down")
+    monkeypatch.setattr(telegram_lotto, "send_urgent_signal", flaky_send)
+
+    agent = LottoAgent()
+    results = [{"signal_id": 1, "fire_level": "urgent"}]
+    ok = await agent._send_urgent_with_retry({"x": 1}, results, task_id="t1")
+    assert ok is True
+    assert attempts["n"] == 3
+
+
+@pytest.mark.asyncio
+async def test_urgent_send_all_fail_returns_false_no_raise(monkeypatch):
+    """urgent 발송이 끝까지 실패해도 raise하지 않고 False (시그널 평가/태스크 보존)."""
+    from app.agents.lotto import LottoAgent
+    from app.notifiers import telegram_lotto
+    import app.agents.lotto as lotto_mod
+    monkeypatch.setattr(lotto_mod, "URGENT_SEND_RETRY_SEC", 0)
+
+    async def always_fail(_event):
+        raise RuntimeError("telegram down")
+    monkeypatch.setattr(telegram_lotto, "send_urgent_signal", always_fail)
+
+    agent = LottoAgent()
+    ok = await agent._send_urgent_with_retry(
+        {"x": 1}, [{"signal_id": 1, "fire_level": "urgent"}], task_id="t1")
+    assert ok is False
+
+
 @pytest.mark.asyncio
 async def test_run_daily_digest_creates_task(monkeypatch):
     """run_daily_digest이 agent_tasks에 task 생성 + result_data 저장."""