Github user NicoK commented on the issue:
https://github.com/apache/flink/pull/4559
Regarding b): I just checked with #4552 and it looks as if the `backlog` is
increased from the `ResultWriter` / task thread, and then both read and
decreased from netty's IO thread. Therefore, we'll not only have a visibility
problem here, but also a synchronisation problem with the increment/decrement,
won't we? ---
