I've dealt with the case where we don't record merge tickets if we haven't seen something new from the remote replica. What we need to deal with next is recognizing when a remote replica is a clone of one we've seen before without reading all the remote changesets
for SD