If a connection is lost and a DNS change is made before an attempt is made to reconnect, the reconnection attempt will fail, because the previous resolution information was cached by the initiator and will be re-used.

I'll be submitting a GitHub pull request for a change that invalidates the cached connection information when a connection attempt fails. This will have no impact on normal operations. The first reconnect attempt will use the cached information, but if it fails, the host name will be resolved again before the next attempt.