Hi, We have experienced several situations where a LeaderSelector does not become eligible to be leaders after being disconnected from zookeeper, even though we periodically call requeue() on it.
This happens because the isQueued-flag is not cleared in all cases. Specifically when a leaderselector is waiting on the mutex. I have submitted a proposal for a patch and a test case on the issue, and would like feedback on it. https://issues.apache.org/jira/browse/CURATOR-73 -- Henrik Nordvik
