[exim] unusually long delays during DATA part of smtp commun…

Hey Guys,

Am having a really bizzare problem here causing us massive grief and
i'm having a hard time trying to replicate/isolate the cause am
hoping you might have a few ideas.

On our primary mail server here i'm noticing major (up to 1-2
minutes) delays on incoming SMTP connections at the end of the DATA
phase. After the final period exim appears to be hanging a majorly
long time before notifying the client that their message is
accepted. The message is eventually accepted but the clients are
timing out, which for our customers is resulting in messages being
held in their outboxes and re-sent multiple times (MUA thinks it
failed, however it just timed out before receiving the OK). This
also seems to result in a stack of TCP connections to exim hung in
FIN_WAIT.

Initially I suspected it may have been Sophos introducing the delays,
disabled it yesterday and the problem seemed to go away, however
today it is happening again and i'm at a loss as to what is causing
the delays.

The machine isn't loaded in any way and has been running solidly for
months. The logs reveal nothing, and what makes it so much harder is
it's next to impossible to replicate in that some incoming sessions
run fine, others don't and this isn't dependent on the content of the
mail or anything - I can do 10 manual sessions telnetting to
localhost and some will work fine, some with hang with exactly the
same sequence of commands.

If anyone can suggest any techniques that might help to diagnose this
i'd be very appreciative.

Thanks,
Marcus