At Summit 23 aerozol was tasked with gathering a list of what methods other sites use to combat spam, e.g. Discourse's auto spam-flagging feature*, which could be implemented in SpamBrainz.

Discourse (main source):

• User typed post suspiciously fast
• You can create 'watched words', automatic actions based on post content
• Limit the number of links new users can make to an outside domain while they are new (defaults to 3)
• Limit the number of new accounts that can be made from any given IP address (defaults to 3)
• Flag sock puppets checkbox - When a user replies to a post by a user from the same IP address, mark both accounts as potential spam
• Other forum discussions mentioned intermittent high-volume region-based spam being effectively blocked with code like

  {{auto block first post regex : \p{Hangul}{3}}}
  

  (If a first post contains more than 3 Korean letters, send it to the 'require approval' queue)

Other

• IPs where a certain amount of spam/blocked users have originated from
• ListenBrainz listens on a loop for x times, or users submitting listens without ever 'sleeping'

Third party anti-spam tools
The overwhelming amount of spam seems to be fought using third party tools, in particular Akismet, who spend a organisations-worth of effort keeping a step ahead.
After digging through a lot of search results I've come to the conclusion that finding a way to route new user profile text (or even text written by that user anywhere, in any project) through one of these tools would by far be the most effective way to stay on top of spam. Including in terms of our time investment (and, famously, time = money).

*This Discourse flag feature is quite useful, as it lets users post/take actions, but holds them all until a moderator OK's them. This discourages spam editors, but doesn't make new users that are falsely falgged wait before taking actions/posting, which can cause them to not bother and move on.