Currently, when trying to make a cluster in replicaset mode healthy, we rerun the entire connection process.

This means we:

1) break existing connections, that may not be necessary
2) discard ping time information

Both are not ideal and should be fixed.