There are two issues:

1. A router checker ping can timeout, causing the mdh to be invalidated. We need to recreate the mdh in that case
2. When re-transmitting a message, even if the peer is marked as down we should re-transmit the message to fulfill it's retry quota.