Despite the message subject I wouldn't say your SA install is *very* broken,
only unmaintained and somewhat broken. One would expect that an
unmaintained filter would let more spam through with time. SPam patterns
change as the spammers get better at avoiding old blocks, so we have to
invent new blocks.
What I see from your headers is that you don't seem to have any third party
rules such as from RulesEmporium, and it looks like your Bayes database is
at least partially messed up - it rated that second spam as pure ham, and
was completely unsure in the first case.
I'd suggest the following:
1. Upgrade to 3.1 or 3.0.5.
2. If you have auto-learning enabled, adjust the ham threshold downward.
3. Blow away the bayes database and start over.
4. Get some SARE rules and perhaps some of the old standards like
chickenpox, which still work pretty well.
With Bayes, I'd recommend turning off auto-leanring initially and
hand-feeding it the required 200+ ham and ideally about a thousand spam so
it gets a good start in life. Then you can turn auto-learning back on
reasonably safely.
Also check your trusted networks and internal networks settings. I don't
see problems in what you posted, but these have been a problem for a great
many people, and their importance isn't sufficiently empahsised in the
install documentation.
Loren
