Bounding Shutdown Retries By Attempt Count Strands The Transcript
By Rokas Remeika. Published 24 October 2026. When a voice agent worker shuts down, the framework gives it a fixed time window before killing the process. If your transcript persistence logic retries based on a maximum attempt count, the combined HTTP timeouts can exceed the shutdown window. The process dies mid-write, losing data that only existed in memory. The fix is to budget retries by wall…
The article discusses a bug in a voice agent framework where transcripts and end-of-call records are lost during shutdown due to an incorrect retry logic approach. When a worker process shuts down, the framework provides a fixed window for cleaning up and saving state, after which the process is killed. The issue arises when persistence logic retries by attempt count rather than wall clock time.
If the combined HTTP timeouts exceed the shutdown window, the process can be killed mid-write, losing data in memory. The fix is to budget retries by wall clock time, ensuring the budget plus one maximum-length HTTP timeout remains within the framework's shutdown window.
Brief written by urgent.news from Dev.to's own syndicated text. Machine-written — may contain errors; check the original before relying on it.
