The filter is this goal, run on the hub:
Status is derived at read time from the hub's probes: a fresh success is up; one missed probe after a good one is deemed unreachable; a node that is alive but draining is maintenance; and anything that remains unreachable long enough is down.