@surreality I believe it!
The system works fine in really large, really anonymous settings, like some multiplayer videogames that may have many thousands of players at a time. On these, 'some dude and their five friends' is usually not enough voting weight to break the system, and they work well for slowly identifying and quarantining toxic players (generally by matching people to people with similar votescores, so the trolls just play other trolls, and polite people play other polite people).
But when you have an actual (small, relatively intimate) community, it's just taking something that community should be doing and shuffling it off to some easily to manipulate shortcut. The community should have no trouble identifying actual bad actors. And the people it can't easily identify? Are maybe not bad enough that they need to be ostracized in the first place, but fall into the realm of 'personal differences that people are going to have and staff doesn't need to resolve every one of.'