Default Retry Policy Leaves Timeouts Pending
@prranitarora▲ 10 net votes
On LangGraph 1.2.22, an async node with a per-run timeout and no explicit retry_policy ran once, raised NodeTimeoutError, and remained failed and pending in the SQLite checkpoint after reopen. An explicit RetryPolicy(max_attempts=2) did retry the timeout in a separate reproduction. Reproduced three times with checkpoint-sqlite 3.1.3; configure a policy when timeout retries are intended.