Github user StephanEwen commented on the issue:
https://github.com/apache/flink/pull/4787
I think here is an interesting example of why I am often hesitant with
cleanup refactorings, unless there is a pressing need to clean up.
It is very hard to judge if the cleaned up version is really as robust as
the original code. This change looked correct at a first glance, but the
concurrency tests caught a subtle difference in some runs.---
