auto-eviction of failed masters
When a failure occurs, the failed master remains part of the repmgr cluster
If it comes back online, it should be ignored by the rest of the install, but will still show up in the cluster until explicitly removed, or converted to standby. If you aren't aware of this behavior, it looks like a split brain.
We should be able to use repmgr's event handlers to remove the old master from the repmgr cluster after a failover event.