By delaying the replay for a second, we give the monitor a small chance to adap to master failures. It'll also prevent rapid re-querying if multiple transaction replays are supported.