Github user anmolnar commented on the issue:
https://github.com/apache/zookeeper/pull/119
@hanm @arshadmohammad This PR hasn't been updated for ages unfortunately,
but looks like it indicates a pretty significant bug in Zk client library.
Additionally we could also fix an long outstanding flaky test by fixing this,
so killing two birds with a single shot.
Although Akihiro's comment on the jira sounds reasonable: by using TCP
there's no such thing as "packet loss" and we should better try to find the bug
in Zookeeper code, I believe that waiting indefinitely is generally a bad thing
and introducing a reasonable and configurable timeout would be beneficial.
Introducing such thing in 3.6 is no harm and I would consider it merging to
3.5 too.
---
