Thanks for posting this. From what I've found online, this seems to be a
problem related to Akka. For example there is this PR to fix this:
https://github.com/akka/akka/issues/16623
Can you check whether this should have been fixed in the Akka version we are
using and post to the Akka ML if so?
While experimenting in a cluster setting I was experiencing some hardware
failures causing some taskmanagers to be unregistered and as a result also
failing my streaming jobs. In the logs after the taskmanager dies I see
some akka Exceptions. I think they are harmless compared to loosing
taskmanage
