Description
I'm 99% certain this is a test artifact, I think I see the problem. It'll take me a lot of beasting to nail it though.
Working hypothesis is that the when we test for whether the new searcher has no messages, we can test for no messages being logged against the watcher before the new one really gets active.