I have posted the details of my issue to the SpamAssassin User Group:
In short, email messages are coming through with very low X-Spam-Score header values.
I dumped a few sample emails and fed them to "spamassassin" on the command line and the scores were much, much higher; they seemed "correct".
Folks on the SA User Group suggested that I check whether network tests are being run when SA is called via AMaViS.
The problem is that I don't know how to call AMaViS directly, e.g., from the command line, in a way that simulates how the message will be processed in a real situation.
I did check my AMAViS configuration, and on the surface, network tests appear to be enabled:
# grep -ir sa_local_tests_only /etc/amavis
/etc/amavis/conf.d/20-debian_defaults:$sa_local_tests_only = 0; # only tests which do not require internet access?
Does ISPConfig somehow disable network tests by some other means?
I did have a look at http://www.ijs.si/software/amavisd/#faq-spam
and walk through the items in the section entitled SpamAssassin returns different score or null score or triggers different set of SA rules when called from amavisd-new, as compared to the command-line utility spamassassin on the same message. What is wrong?
In my tests, I am calling "spamassassin" from the "amavis" user account, amavis is not running in a chroot
ed environment, and nothing else that is mentioned stands-out. My Bayes database looks fine, and I don't think it's a factor here.
I realize that ISPConfig allows for SPAM policies to be applied at the domain and/or mailbox level, so, ideally, I'd like to be able to "simulate" the process by which email messages for a specific domain/mailbox will be scored when fed to SA via AMaViS.
Thanks in advance!