I notice that in many of the spam messages that sneak through, the X-Spam-Status header does not have any BAYES_* score. Does this indicate a problem? Should all messages includes a BAYES score?
I am using a single bayes database for our whole network, mounted via NFS. But it seems to lock it for several seconds for each user's access to it, so I could see that bayes access might often time out while waiting for some other's lock to expire? If my theory is correct, does this mean I need to move to having the bayes database in an SQL database? Thanks...
