Problem with fetchmail, postfix, and DynDNS SMTP forwarding

[depontius]

I have a vanity domain registered through DynDNS, along with their inbound and outbound forwarding services. Incoming email goes to gmail, which I then grab with fetchmail. Outbound email goes to DynDNS SMTP forwarding. This all works - usually.

Every now and then someone sends me an email with a badly oversized attachment. When that happens, the whole system goes kerflooey. What's worse, it does so relatively silently, so it may go on for a few days until I discover the problem. As it did this week when a cousin tried to send an email with some 20+ MB of attached photos.

When I get the oversized emails, something bad seems to happen between fetchmail and postfix, and I still don't have a clear picture of exactly what has happened. Usually by the time I discover the problem my postfix queue has hundreds of deferred emails, my DynDNS SMTP outbound quota has been used up for days, and I'm more focused on restoring service than on forensics. By the time I can start thinking about finding out what happened, I've destroyed the evidence in the process of getting things running again. Like I just did a few hours ago.

Has someone had experience with a situation like this, and can give me any pointers?

That lacking, I guess I can try sending an oversized attachment from my work email, watch the bad stuff start to happen, then access gmail directly and read the bad email there (thereby making it so fetchmail won't try to read it, again) before things get out of hand. At that point I should have enough simple evidence to have some hope of figuring out how to stop this.

But I know I'm stumbling about, and am open to suggestions/insights.
_________________
.sigs waste space and bandwidth