This morning I made a mistake with spamassassin. I manually learnt (/usr/bin/sa-learn --spam) an IMPORTANT message as SPAM. I immediately learnt it as HAM, but is that sufficient?
Yes.. SA uses the bayes_seen database to understand it needs to flip completely.
Do I have to delete all of the bayes tokens I accumulated for over 1 year?
If you want to delete them *all* use sa-learn --clear, or if your version doesn't support that, rm -f ~/spamassassin/bayes_*.
That will wipe out your bayes db completely.
