fix(auxiliary): evict async wrappers on poisoned client (follow-up to #23482)

#23482 fixed cache poisoning in the sync path: when a Codex auxiliary
timeout closes the underlying OpenAI client, _evict_cached_client_instance
walks CodexAuxiliaryClient wrappers via their _real_client attribute and
drops the cache entry so the next aux call rebuilds.

The cache key includes async_mode (see _client_cache_key), so the sync and
async clients for the same provider live in two distinct entries pointing
at the same underlying transport. The fix walked the sync wrapper's
_real_client correctly but the async wrappers
(AsyncCodexAuxiliaryClient, AsyncAnthropicAuxiliaryClient,
AsyncGeminiNativeClient) never exposed _real_client at all, so the async
entry survived eviction and kept handing out the poisoned client.

Effect on async aux callers: one timeout now poisons every subsequent
async aux call (compression, vision, session_search, title_generation)
with 'Connection error' until gateway restart -- even while the sync
route recovered as designed in #23482.

Mirror the sync wrapper's _real_client onto each async wrapper so the
existing eviction helper finds them. Three changes, one per wrapper:

- AsyncCodexAuxiliaryClient: self._real_client = sync_wrapper._real_client
  (the underlying OpenAI client)
- AsyncAnthropicAuxiliaryClient: same shape
- AsyncGeminiNativeClient: self._real_client = sync_client (Gemini's
  native facade is itself the leaf; no OpenAI client beneath it)

Update _evict_cached_client_instance docstring to reflect that it now
covers both sync and async wrappers via the same attribute walk.

Test: TestAuxiliaryClientPoisonedCacheEviction.test_evict_cached_client_instance_walks_async_wrapper
seeds both sync and async cache entries pointing at the same leaf and
asserts both are dropped on a single eviction call. Verified the test
fails without the wrapper changes ("async cache entry survived
eviction -- wrapper is missing _real_client") and passes with them.

Refs #23482, #23432

agent/auxiliary_client.py

@@ -900,6 +900,14 @@ class AsyncCodexAuxiliaryClient:
         self.chat = _AsyncCodexChatShim(async_adapter)
         self.api_key = sync_wrapper.api_key
         self.base_url = sync_wrapper.base_url
+        # Mirror the sync wrapper's _real_client so cache eviction by leaf
+        # OpenAI client (e.g. _close_client_on_timeout in #23482) drops
+        # this async entry too. Without this, sync and async cache entries
+        # diverge on poisoning: the sync entry is evicted but the async
+        # entry keeps reusing the closed transport, failing every
+        # subsequent async aux call with 'Connection error' until the
+        # gateway restarts.
+        self._real_client = sync_wrapper._real_client
 
 
 class _AnthropicCompletionsAdapter:
@@ -1035,6 +1043,9 @@ class AsyncAnthropicAuxiliaryClient:
         self.chat = _AsyncAnthropicChatShim(async_adapter)
         self.api_key = sync_wrapper.api_key
         self.base_url = sync_wrapper.base_url
+        # See AsyncCodexAuxiliaryClient: mirror _real_client so cache
+        # eviction on a poisoned underlying client also drops this entry.
+        self._real_client = sync_wrapper._real_client
 
 
 def _endpoint_speaks_anthropic_messages(base_url: str) -> bool:
@@ -2108,9 +2119,13 @@ def _evict_cached_client_instance(target: Any) -> bool:
     transport after a timeout, broken streaming session, etc.) so the next
     auxiliary call rebuilds rather than reusing the dead instance.
 
-    Walks ``CodexAuxiliaryClient`` wrappers via their ``_real_client`` so a
-    timeout that closes the underlying ``OpenAI`` client also evicts the
-    Codex shim that exposed it.
+    Walks both sync and async wrappers (``CodexAuxiliaryClient``,
+    ``AnthropicAuxiliaryClient``, ``AsyncCodexAuxiliaryClient``, etc.) via
+    their ``_real_client`` attribute so a timeout that closes the underlying
+    ``OpenAI`` (or native provider) client evicts every cached shim that
+    exposed it. Async wrappers must mirror their sync sibling's
+    ``_real_client`` for this to work — otherwise the sync entry is evicted
+    but the async entry survives and keeps reusing the dead transport.
 
     Returns True when at least one entry was evicted.
     """

agent/gemini_native_adapter.py

@@ -945,6 +945,12 @@ class AsyncGeminiNativeClient:
         self.api_key = sync_client.api_key
         self.base_url = sync_client.base_url
         self.chat = _AsyncGeminiChatNamespace(self)
+        # Expose the underlying sync client as _real_client so the auxiliary
+        # cache's eviction-by-leaf-client helper (#23482) can find and drop
+        # this async entry when the sync GeminiNativeClient is poisoned.
+        # GeminiNativeClient is itself the leaf (no OpenAI client beneath
+        # it), so we point at the sync_client directly.
+        self._real_client = sync_client
 
     async def _create_chat_completion(self, **kwargs: Any) -> Any:
         stream = bool(kwargs.get("stream"))

tests/agent/test_auxiliary_client.py

@@ -2183,6 +2183,42 @@ class TestAuxiliaryClientPoisonedCacheEviction:
         assert _evict_cached_client_instance(None) is False
         assert _evict_cached_client_instance(MagicMock()) is False
 
+    def test_evict_cached_client_instance_walks_async_wrapper(self):
+        """async_mode is part of the cache key so sync and async share the same
+        underlying OpenAI client across two distinct cache entries. A single
+        timeout that closes the leaf must evict BOTH — otherwise the async
+        entry survives, keeps reusing the dead transport, and every async
+        aux call (compression, vision, session_search) fails fast with
+        'Connection error' until gateway restart even while the sync route
+        recovers.
+
+        Regression for the async-side gap left by #23482, which fixed the
+        sync wrapper's _real_client walk but missed the async wrappers.
+        """
+        from agent.auxiliary_client import (
+            _client_cache, _client_cache_lock, _evict_cached_client_instance,
+            CodexAuxiliaryClient, AsyncCodexAuxiliaryClient,
+        )
+
+        real = SimpleNamespace(api_key="k", base_url="https://chatgpt.com/backend-api/codex",
+                               responses=SimpleNamespace(stream=lambda **k: None),
+                               close=lambda: None)
+        sync_wrapper = CodexAuxiliaryClient(real, "gpt-5.5")
+        async_wrapper = AsyncCodexAuxiliaryClient(sync_wrapper)
+        with _client_cache_lock:
+            _client_cache.clear()
+            _client_cache[("openai-codex", False, None, None, None)] = (sync_wrapper, "gpt-5.5", None)
+            _client_cache[("openai-codex", True, None, None, None)] = (async_wrapper, "gpt-5.5", None)
+        try:
+            assert _evict_cached_client_instance(real) is True
+            assert ("openai-codex", False, None, None, None) not in _client_cache
+            assert ("openai-codex", True, None, None, None) not in _client_cache, (
+                "async cache entry survived eviction — wrapper is missing _real_client"
+            )
+        finally:
+            with _client_cache_lock:
+                _client_cache.clear()
+
     def test_codex_timeout_evicts_cached_wrapper(self):
         """The timeout closer evicts the cache entry that wraps the closed client."""
         from agent.auxiliary_client import (