flink_task_checkpointalignmenttime

Rising checkpoint failures indicate upstream backpressure, state growth issues, or resource exhaustion that will eventually cause job restarts and data loss.

Slow RocksDB flushes cause write stalls that propagate upstream as backpressure, degrading throughput and increasing checkpoint durations.