When we shutdown the replica set monitor, all we do is shutdown and join it's task executor. We should also fail all requests to managed rsms to make sure any callers that are relying on things timing out actually timeout (and don't hang)