The code has been refactored so much since this bug was reported, I'm not even sure the originally reported bug is still present.
What used to happen was: assuming no dns entries were present, lookups for a s2s connection used to check:
transport.example.co.uk, then
example.co.uk, then
co.uk, then
uk
Now there's no where in the xmpp rfc's that say that if a dns lookup fails that you should strip the left most part and re-try. If the behaviour in trunk today is to fail after the first dns lookup, then I think we can just close this ticket as 'fixed-due-to-previous-refactoring'.
BTW, the relevant code is in OutgoingServerSession around line 160.