### Check for existing issues

- [x] I have searched the existing issues and checked that my issue is not a duplicate.

### What happened?

The synchronous `Logging.success_handler` and `Logging.failure_handler` (`litellm/litellm_core_utils/litellm_logging.py`) iterate the live `self.model_call_details` dict directly (`for k, v in self.model_call_details.items():`) while building the kwargs passed to sync callbacks (langfuse, logfire, greenscale, athina, traceloop).

These sync handlers run on a **worker thread** (dispatched via `executor.submit` / `threading.Thread`), while `async_success_handler` / `async_failure_handler` run on the **event loop** and insert **new keys** into the *same* `self.model_call_details` object (e.g. `async_complete_streaming_response`, `response_cost`, `standard_logging_object`, and — on failures — `log_event_type`, `exception`, `traceback_exception`, `end_time`). When the async path inserts a key while the sync loop is mid-iteration, Python raises `RuntimeError: dictionary changed size during iteration`, which aborts that callback's logging for the request (the exception is swallowed by the handler's non-blocking try/except, so the log is silently dropp…