Which metrics are useful for monitoring?

Obviously `unbound_up` is useful for monitoring in the sense that alerting can alert to a non-zero or absent value.

For the other exposed metrics, it's not clear what might indicate some sort of problem requiring an alert.

Anyone got a useful set of alerting rules they use?