With spamassassin and amavis, I know how to use sa-learn to dump the magic tokens from the bayes DB to get an idea of the number of tokens learned and whether learning an email as either ham or spam has had any effect.
How would I go about checking the same with mailcow? I can’t shake the feeling that the mails I tell rspamd to learn as ham are not having any effect on my bayes db.
Slightly off-topic, but I am also occasionally seeing this error message when attempting to learn emails from the quarantine as spam:
Fuzzy hash learn error: Message sending error
What could cause this and how can I investigate what happened in such a case?