Sophisticated spam bot detection

[kangalio]

Currently, spam bots are dealt with using a blacklist of words whose mention gets anyone banned instantly. This is a suboptimal approach, because

1. it prevents server members from genuinely discussing about the blacklisted topics
2. real members accidentally mentioning one such word don't get any explanation or appeal possibility

Here's the Discord message that sparked this discussion: https://discord.com/channels/273534239310479360/461038099682689054/787343206660440136

By implementing the spam bot detection in Ferris, more sophisticated detection could be done than what Carl-bot (aka Turtle) supports. To be precise, members mentioning the "forbidden" words won't be banned if they have sent a certain number of messages. This is to distinguish between genuine human server members and people/bots who just joined to spam.

[kangalio]

Is this sort of sophisticated spam bot detection desired? Is the proposed behavior (members over a certain number of messages won't be banned from forbidden words) good, or is there a better alternative? And even, is spam bot detection still required? It seems like there is much less spam bot traffic on the server than there used to be.