Github user zentol commented on the issue:
https://github.com/apache/flink/pull/4939
yes there could be a sleeper failure in hadoop, but there's nothing we can
do to fix that besides relocation jackson in either hadoop or aws, which is
probably not possible.---
