Deadlock between core reload and IndexFetcher

[Joel Bernstein]

I'll create a jira for this shortly but wanted to mention this on the dev
list. We've run into a deadlock which occurs when a core reload occurs at
the same time the IndexFetcher is running. The effect of this is that
replication stops until Solr is restarted. There is a work around which is
to disable polling before reloading the core but ideally we'd get this
deadlock fixed.