mirror of
https://github.com/langchain-ai/langgraph.git
synced 2026-09-26 19:45:00 +02:00
Sleeping in the worker thread on resume is not scalable: with max_interval=128s and many concurrent graphs, worker threads would be held hostage doing nothing. Instead, only restore the attempt count on resume and let external schedulers (e.g. LangGraph Server) use the persisted retry_ts to delay re-scheduling the task. https://claude.ai/code/session_01768mrKrVCCtXdJgRNFdWrb