Cluster intermittently enters idle state after initial query execution

After initial queries execute successfully, the Trino cluster enters an idle state where no queries process despite resources being available. CPU utilization drops below 30% and memory below 40%. The cluster spontaneously recovers after 2-90 minutes, but may return to idle state. GC is not the cause (no full GC, all cycles under 110ms).

Monitor coordinator logs for query scheduling or dispatch errors. Check query queue depth and coordinator resource utilization. Review network connectivity between coordinator and workers. Examine query.remote-task.max-error-duration (1m) to see if tasks are prematurely failing. Verify metastore connectivity and response times. Check if retry-policy=QUERY and query-retry-attempts=2 are causing retry loops.