diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py
index d3fe1406abd..0f95cda3d46 100644
--- a/litellm/proxy/proxy_server.py
+++ b/litellm/proxy/proxy_server.py
@@ -5565,7 +5565,8 @@ async def async_data_generator(
             request_data=request_data
         )
         model_mismatch_logged = False
-        # Chunks for post-guardrail log: use exclude_none=True only so stream_chunk_builder gets required keys
+        # Chunks for post-guardrail log: use exclude_none=True only so stream_chunk_builder gets required keys.
+        # This list holds the full stream in memory until the background post-guardrail log task completes.
         _streaming_chunks_for_log: List[Dict[str, Any]] = []
         # Use a running string instead of list + join to avoid O(n^2) overhead.
         # Previously "".join(str_so_far_parts) was called every chunk, re-joining
diff --git a/tests/test_litellm/proxy/hooks/test_async_post_guardrail_log_success_event.py b/tests/test_litellm/proxy/hooks/test_async_post_guardrail_log_success_event.py
index 88d94592b08..e66cbd36beb 100644
--- a/tests/test_litellm/proxy/hooks/test_async_post_guardrail_log_success_event.py
+++ b/tests/test_litellm/proxy/hooks/test_async_post_guardrail_log_success_event.py
@@ -85,18 +85,22 @@ async def test_post_guardrail_log_base_custom_logger_not_invoked():
     """Base CustomLogger (no override) is not invoked; only overriders are called."""
     base_logger = CustomLogger()
     overriding_logger = PostGuardrailLogger()
-    # Base first, then override: only overriding should be called
-    with patch("litellm.callbacks", [base_logger, overriding_logger]):
-        from litellm.proxy.utils import ProxyLogging
-        from litellm.caching.caching import DualCache
-
-        proxy_logging = ProxyLogging(user_api_key_cache=DualCache())
-        await proxy_logging.async_post_guardrail_log_success_event(
-            data={"model": "gpt-4"},
-            response=ModelResponse(id="r1", choices=[], model="gpt-4"),
-            user_api_key_dict=UserAPIKeyAuth(api_key="test-key"),
-        )
+    with patch.object(
+        base_logger,
+        "async_post_guardrail_log_success_event",
+        new_callable=AsyncMock,
+    ) as mock_base_method:
+        with patch("litellm.callbacks", [base_logger, overriding_logger]):
+            from litellm.proxy.utils import ProxyLogging
+            from litellm.caching.caching import DualCache
 
+            proxy_logging = ProxyLogging(user_api_key_cache=DualCache())
+            await proxy_logging.async_post_guardrail_log_success_event(
+                data={"model": "gpt-4"},
+                response=ModelResponse(id="r1", choices=[], model="gpt-4"),
+                user_api_key_dict=UserAPIKeyAuth(api_key="test-key"),
+            )
+        mock_base_method.assert_not_called()
     assert overriding_logger.called is True