How do you diagnose and fix blocking work on an asyncio event loop?

Telemetry needs to separate event-loop delay, dependency delay, and worker saturation. A thread offload can make the loop responsive while hiding a growing executor backlog, so I monitor both. For example, if file parsing moves to ten threads, I cap submissions so fifty requests cannot retain fifty large files while waiting. Load tests use realistic concurrency and payload size, and timeouts remain outside the offloaded call because cancellation of the awaiting task does not necessarily stop the underlying thread function.