could use some kind of notification system when nodes go down for long periods of time. otherwise we only know when we decide to pull up the monitor.