How does the "learn as spam/ham" feature in aliases work?

d-u

Hi everyone,
I’ve been testing the “learn as spam” feature for aliases, but I’m having trouble understanding how it contributes to spam learning.
So, I created an alias with the “learn as spam” option enabled, sent mails to this and checked the Rspamd UI but didn’t see any changes to the Bayes or fuzzy classifiers, only the “SPAM_TRAP” symbol is assigned with a 0 score and the mail isn’t in the mailbox.

Upon inspecting the configuration, I noticed that when an email is sent to a “learn as spam” alias, the SPAM_TRAP symbol is added in the dynamic map. However, this symbol doesn’t appear to have a score assigned to it by default, and I couldn’t find any documentation explaining what actions are triggered by it.

So, what is the intended function of the SPAM_TRAP symbol in a default mailcow installation?
Does sending an email to a “learn as spam” alias automatically train the Bayes and/or fuzzy filters?
If so, is there a delay, or is there a specific command I need to run to see the learning take effect?

Am I misunderstanding how this feature is supposed to be used?
(Also I’m a bit confused why it’s called “SPAM_TRAP”, but has, as far as I see, nothing to do with the spamtrap module and can’t be configured by that, as it’s a dynmap given symbol.)

Thanks!