I assume you are still using 2.63? If so, you aren't the first to notice
this.
There seems to be several theories, but I don't know that anyone has ever
come up with a definitive proof of how this comes about. Hopefully it will
be gone on 3.0.
One theory: Mails that score near Bayes-50 won't show a Bayes score. But
I'd think they should still possibly show other scores.
Another theory: If the machine gets really swamped, it seems that messages
will get skipped.
Yet another: If network tests are timing out (or maybe just taking too long)
the processing may get aborted for the message.
Loren
