Comparison

s2smanager: Queue db:verify unless we already sent a db:result (if we had then it can could a dialback deadlock). Also remove some redundant code which could cause a db:result to be sent while still negotiating features (e.g. TLS) and break things. Collectively these fix a 'random' s2s failure (usually with ejabberd for some reason) - resulting in an 'unbound prefix' XML error, or 'ssl handshake failure'. Was this commit message long enough? I think so.