fix - use tp executor (#7509)

Author: ishaan-jaff

litellm/utils.py

@@ -915,9 +915,12 @@ def wrapper(*args, **kwargs):  # noqa: PLR0915
 
             # LOG SUCCESS - handle streaming success logging in the _next_ object, remove `handle_success` once it's deprecated
             verbose_logger.info("Wrapper: Completed Call, calling success_handler")
-            threading.Thread(
-                target=logging_obj.success_handler, args=(result, start_time, end_time)
-            ).start()
+            executor.submit(
+                logging_obj.success_handler,
+                result,
+                start_time,
+                end_time,
+            )
             # RETURN RESULT
             if hasattr(result, "_hidden_params"):
                 result._hidden_params["model_id"] = kwargs.get("model_info", {}).get(
@@ -1122,11 +1125,12 @@ async def wrapper_async(*args, **kwargs):  # noqa: PLR0915
             asyncio.create_task(
                 logging_obj.async_success_handler(result, start_time, end_time)
             )
-            threading.Thread(
-                target=logging_obj.success_handler,
-                args=(result, start_time, end_time),
-            ).start()
-
+            executor.submit(
+                logging_obj.success_handler,
+                result,
+                start_time,
+                end_time,
+            )
             # REBUILD EMBEDDING CACHING
             if (
                 isinstance(result, EmbeddingResponse)