This is a followup to https://issues.redhat.com//browse/OCPBUGS-44694 to make the monitor resilient in all configurations, including things like 5 node control planes. Instead of relying on a longer fall time, we can just let HAProxy report its own ability to reach any backend, which means under ordinary circumstances this check will never fail. There will no longer be any issue with pathologically bad call chains where we happen to hit backends that are down but haven't been detected yet.