[Mailman-Users] Is There a list member limitation?

[Stephen J. Turnbull]

Goodman, William writes:

 > We have a very busy mail server so I set it to 4 hours, when I used the 
 > default we had a mailq of over 200 everyday which slowed our
 > performance.
 > So it will keep trying to send it for 4 hours then it gets bounced.

Sure, but the main point is that the user can't find it in his/her
mailbox because that mailbox is "unavailable".

 > > to=<user at somewhere.com>, relay=mail.somewhere.com[212.137.44.179]:25,
 > > delay=8.3, delays=0.02/4.4/0.67/3.3, dsn=4.0.0, status=deferred (host
 > > mail.somewhere.com[212.137.44.179] said: 451 <user at somewhere.com>...
 > > Requested mail action not taken: mailbox unavailable (in reply to RCPT
 > > TO command)) Jan 13 15:25:00 my-mailserver postfix/smtp[28262]:
 > > B405E2D858F

This looks like greylisting to me.  (Maybe broken; state of the art
greylisters tell you how long to wait before resending.)

Another possibility, especially if "somewhere" == aol or hotmail, is a
tarpit.  They think you may be spamming their members, and they're
forcing you to slow down.

I'll give you one scenario you might think about that connects a
couple of the symptoms you report.

1.  Some corresponding hosts have decided you might be a spammer, and
    are rate limiting you by temporarily failing delivery to any mailbox.

2.  The queue built up not because of load on your server, but because
    those hosts will only let you send to one address every 10 minutes
    or something like that.

3.  When you cut the retry period to 4 hours, you simply arranged to
    bounce mail that would eventually have succeeded in due time, but
    you *must* pay that "due time" to succeed at those destinations.

If it's indeed one or more of the big services, they usually have an
AUP for lists.  You can register with them, provide bona fides (such
DKIM headers in your outgoing posts), and magically things will start
flowing well.  For smaller hosts, you might just negotiate directly
with the postmaster.

Another possibility is that you might want to just accept the
slowdown, but to reduce resources used by your host, instead of doing
the first retry very soon, you could wait several hours.