Github user StephanEwen commented on the issue: https://github.com/apache/flink/pull/5718 Having a custom timeout like @zentol suggests could be helpful in the long run. For now, though, I personally agree that the timeouts are counter productive (I actually never understood what they are supposed to solve). I prefer a deadlocking test to stay around, so I can jstack to the process, attach a debugger, pull a heap dump, whatever. The only environment where we cannot do that is the CI server, which has a timeout and thread dump already...
---