Problem
HTTP watchers wake every 1 second even when no poll is due, and enqueue overlapping REST tasks per application when responses are slow.
Proposed solution
- Poll Flink/Spark apps every 5 seconds in steady state; keep 1-second fast polling only during start/stop operations
- Align Flink cluster watcher schedule with its 30-second interval
- Skip overlapping watch tasks per application with single-flight guards
Problem
HTTP watchers wake every 1 second even when no poll is due, and enqueue overlapping REST tasks per application when responses are slow.
Proposed solution