Hi,
From what i've noticed, when an action in the mailman web interface sends out a message, it does this directly by calling another part of the system that makes the smtp connection. Sending out just one message this way is probably not much of a problem... but if it sends out a large batch of messages (like when you approve a large number of messages), the http connection is staying open untill all of the messages are send to the MTA on the localhost... I was wondering if maybe it could be better if the cgi scripts 'mark' messages as being allowed to post to the list, and a mailman cron job picking them up next time... i think this could also make the UID-thing a bit less complicated as the cgi scripts don't have to send out any mail, just make a small change in the database. And maybe this also can prevent the high system loads when the cgi scripts are feeding messages to the MTA and the MTA is trying to deliver most of them at the same time (at least, that's what's happening in my mailman/exim configuration)
maybe i'm completely wrong, but these were just my thoughts :)
Ricardo.
--