Brad Knowles wrote:
This is a postfix problem. Find the sites that cause the problems, and set up separate queues for them, with different parallelism and re-queueing/re-try schedules.
Normally, I would defer to you for all things MTA as your knowledge and experience in this area are far greater than mine, but I have been under the impression that if an MTA has two or more messages to deliver to the the same remote MX, that it can and perhaps will deliver them in one SMTP session as follows
connect HELO (or EHLO) MAIL FROM RCPT TO DATA . MAIL FROM RCPT TO DATA . ... QUIT
I.e., each message will be sent with a separate MAIL FROM, RCPT TO and DATA, but more than one message will be sent within a single connect, HELO, QUIT session.
If this is actually done (or can be done) by Postfix, it might help Fil, since the complaint in Fil's post was
(host mx01.xxxx.xxx[xxxxx] refused to talk to me: 421 4.3.2 Connection rate limit exceeded.)
in response to a connect request. Thus, if more messages could be sent over one connection, that would reduce the connection rate even though it wouldn't reduce the message rate.
Also, I think from earlier messages in this thread (http://mail.python.org/pipermail/mailman-users/2009-January/064588.html et seq) that Fil is already successfully using Postfix VERPing with Mailman, so the only question here is can Postfix be convinced to send multiple messages to the same MX over a single connection.