I have had instances where a machine gets removed from a site, but I don't receive any alerts. Looking at the available ArcGIS Monitor metrics, I don't see one that addresses this situation. (The recent update might help though.) ArcGIS Monitor reports on the number of requests received by an ArcGIS Server site, but the total number does not tell me about the health of an individual machine in the site. If the number of requests received decreases on an individual machine, knowing this would help me know that something has gone wrong with the load balancer or an individual machine. My idea is to add more ArcGIS Server metrics for the activity of an individual machine, and "requests received" would be a very helpful metric.