fix(retry): route non-OpenAI usage-limit errors to the resumable stop

Follow-up to the usage-limit terminal classification (review feedback): retry
suppression is provider-agnostic (text-marker based), but run_strix_scan only
mapped openai.RateLimitError to the resumable stop. A LiteLLM-routed provider
that surfaces a usage-limit as a different exception type therefore reached the
generic failure path (root marked failed, re-raised) instead of the promised
resumable stop with the `strix --resume` hint.

Route any error where codex.is_usage_limit_error(exc) is true to the same
resumable stop as RateLimitError. Adds a regression test covering a non-OpenAI
(LiteLLM-routed) usage-limit exception.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

strix/core/runner.py

@@ -18,7 +18,7 @@ from openai import RateLimitError
 
 from strix.agents.factory import build_strix_agent, make_child_factory
 from strix.agents.prompt import render_system_prompt
-from strix.config import load_settings
+from strix.config import codex, load_settings
 from strix.config.models import (
     StrixProvider,
     configure_sdk_model_defaults,
@@ -521,24 +521,36 @@ async def run_strix_scan(
             with contextlib.suppress(Exception):
                 await coordinator.set_status(root_id, "stopped")
         return None
-    except RateLimitError as exc:
-        logger.warning(
-            "Scan %s stopped: persistent rate limit from the LLM provider (%s). "
-            "Resume with 'strix --resume %s' once the limit clears.",
-            scan_id,
-            exc,
-            scan_id,
-        )
-        if root_id is not None:
-            with contextlib.suppress(Exception):
-                await coordinator.set_status(root_id, "stopped")
-        return None
     except (asyncio.CancelledError, KeyboardInterrupt):
         logger.info("Scan %s interrupted by the user", scan_id)
         if root_id is not None:
             with contextlib.suppress(Exception):
                 await coordinator.set_status(root_id, "running")
         raise
+    except Exception as exc:
+        # A usage-limit rejection is resumable regardless of the provider's
+        # exception class: OpenAI raises RateLimitError, but a LiteLLM-routed
+        # provider may surface the same exhausted-window error as a different
+        # type. Both retry layers already stop retrying it (see
+        # codex.is_usage_limit_error), so route it to the same resumable stop
+        # instead of the generic failure path.
+        if isinstance(exc, RateLimitError) or codex.is_usage_limit_error(exc):
+            logger.warning(
+                "Scan %s stopped: persistent rate limit from the LLM provider (%s). "
+                "Resume with 'strix --resume %s' once the limit clears.",
+                scan_id,
+                exc,
+                scan_id,
+            )
+            if root_id is not None:
+                with contextlib.suppress(Exception):
+                    await coordinator.set_status(root_id, "stopped")
+            return None
+        logger.exception("Strix scan %s failed", scan_id)
+        if root_id is not None:
+            with contextlib.suppress(Exception):
+                await coordinator.set_status(root_id, "failed")
+        raise
     except BaseException:
         logger.exception("Strix scan %s failed", scan_id)
         if root_id is not None:

tests/test_runner_rate_limit.py

@@ -1,4 +1,4 @@
-"""Tests for graceful handling of persistent RateLimitError in run_strix_scan."""
+"""Tests for graceful handling of persistent usage/rate-limit errors in run_strix_scan."""
 
 from __future__ import annotations
 
@@ -24,11 +24,22 @@ def _make_rate_limit_error() -> RateLimitError:
     return RateLimitError("rate limited", response=response, body=None)
 
 
-@pytest.mark.asyncio
-async def test_persistent_rate_limit_stops_gracefully(
-    monkeypatch: pytest.MonkeyPatch, tmp_path: Any, caplog: pytest.LogCaptureFixture
-) -> None:
-    """A persistent RateLimitError stops the scan (root -> 'stopped') without raising."""
+class _ProviderUsageLimitError(Exception):
+    """A non-OpenAI (e.g. LiteLLM-routed) provider exception whose text carries the
+    usage-limit marker but which is *not* an ``openai.RateLimitError``."""
+
+
+def _make_litellm_usage_limit_error() -> Exception:
+    return _ProviderUsageLimitError(
+        "litellm.RateLimitError: 429 - {'error': {'type': 'usage_limit_reached', "
+        "'message': 'The usage limit has been reached'}}"
+    )
+
+
+async def _run_scan_raising(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Any, error: BaseException
+) -> tuple[Any, AgentCoordinator]:
+    """Drive run_strix_scan with a fully mocked scan whose agent loop raises ``error``."""
     monkeypatch.setattr(runner, "run_dir_for", lambda _scan_id: tmp_path)
     monkeypatch.setattr(runner, "runtime_state_dir", lambda _run_dir: tmp_path)
     monkeypatch.setattr(runner, "setup_scan_logging", lambda _run_dir: lambda: None)
@@ -70,21 +81,24 @@ async def test_persistent_rate_limit_stops_gracefully(
     monkeypatch.setattr(runner, "make_child_factory", lambda **_kwargs: lambda **_k: object())
     monkeypatch.setattr(runner, "open_agent_session", lambda _root_id, _db: object())
 
-    async def _raise_rate_limit(*_args: Any, **_kwargs: Any) -> None:
-        raise _make_rate_limit_error()
+    async def _raise(*_args: Any, **_kwargs: Any) -> None:
+        raise error
 
-    monkeypatch.setattr(runner, "run_agent_loop", _raise_rate_limit)
+    monkeypatch.setattr(runner, "run_agent_loop", _raise)
 
     coordinator = AgentCoordinator()
+    result = await runner.run_strix_scan(
+        scan_config={"targets": [], "scan_mode": "deep"},
+        scan_id="scan-test",
+        image="img",
+        coordinator=coordinator,
+    )
+    return result, coordinator
 
-    with caplog.at_level(logging.WARNING):
-        result = await runner.run_strix_scan(
-            scan_config={"targets": [], "scan_mode": "deep"},
-            scan_id="scan-test",
-            image="img",
-            coordinator=coordinator,
-        )
 
+def _assert_stopped_resumably(
+    result: Any, coordinator: AgentCoordinator, caplog: pytest.LogCaptureFixture
+) -> None:
     assert result is None
     root_ids = [aid for aid, parent in coordinator.parent_of.items() if parent is None]
     assert len(root_ids) == 1
@@ -92,3 +106,29 @@ async def test_persistent_rate_limit_stops_gracefully(
     # the resume hint must carry the real scan id, not a literal placeholder
     assert "strix --resume scan-test" in caplog.text
     assert "<run_name>" not in caplog.text
+
+
+@pytest.mark.asyncio
+async def test_persistent_rate_limit_stops_gracefully(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Any, caplog: pytest.LogCaptureFixture
+) -> None:
+    """A persistent OpenAI RateLimitError stops the scan (root -> 'stopped') without raising."""
+    with caplog.at_level(logging.WARNING):
+        result, coordinator = await _run_scan_raising(
+            monkeypatch, tmp_path, _make_rate_limit_error()
+        )
+    _assert_stopped_resumably(result, coordinator, caplog)
+
+
+@pytest.mark.asyncio
+async def test_persistent_usage_limit_non_openai_stops_gracefully(
+    monkeypatch: pytest.MonkeyPatch, tmp_path: Any, caplog: pytest.LogCaptureFixture
+) -> None:
+    """A usage-limit error from a non-OpenAI (LiteLLM-routed) provider — not an
+    ``openai.RateLimitError`` — must also land on the resumable stop path, not the
+    generic failure path."""
+    with caplog.at_level(logging.WARNING):
+        result, coordinator = await _run_scan_raising(
+            monkeypatch, tmp_path, _make_litellm_usage_limit_error()
+        )
+    _assert_stopped_resumably(result, coordinator, caplog)