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.
---