Github user kl0u commented on the issue:
https://github.com/apache/flink/pull/4900
In this case, the original problem will persist, right? (The one in the
JIRA).
In addition, given that:
1) we check the status of the service before firing a timer, and
2) the status is set (through `quiesce()`) in the same guarded block as the
close of the operator, so no timers can fire in the meantime,
I am not sure if we could see any negative side-effects from this change.
What do you think?---
