LISTSERV was not designed from the ground up as a spam filtering
clearinghouse that can be connected to arbitrarily slow spam filters. It was
designed as a high-performance e-mail list manager :-) When viruses came, a
virus scanning function was added, and there was no noticeable performance
impact. Scanning the message twice is in practice necessary because:

1. If the message contains a virus, processing should be aborted as soon as
reasonably possible. The message should not be held in the spool waiting for
something to happen, etc. For this reason, the message should be scanned as
early as reasonably possible, considering that this is a per-list setting.

2. In THEORY, if the full entire message did not contain any viruses, then
any subset of that message will also be virus-free. But in practice this is
not guaranteed, since virus scanners use a lot of heuristics and virus
writers spend a lot of energy trying to defeat them. Given the very low cost
of virus scans, it would be hard to justify not scanning the final message
that is actually going to be sent to thousands of people whose computer
might be at risk if you decided that the first scan was enough.

The spam filter is implemented as an extension of the virus scanner, and a
side effect is that spam scanning is done twice in case the message is
rewritten (which is not very common). This is bad because it takes longer,
and it is good because sometimes when you run a message twice through the
spam scanner, it will go through the first time and be blocked the second
time, because of network delays (in fact, at one point we used a spam exit
that ran the query twice through SpamAssassin and used the highest of the
two numbers, which gave better results, but it turned out to be too slow in
practice). Either way, this is no big deal as long as the spam filter is
reasonably fast. Unfortunately, the most popular spam filter happens to be
extremely slow.

  Eric