Wouldn't a detection system be way better? I can see a machine learning model handling this rather well. Correlate the main accounts to their upvoters across all their posts and create a flag if it returns positive. It would be more of a mod tool, really.
I have already ran into a very obvious Russian troll factory account and it really drags down the quality of the place. Freedom of speech shouldn't extend to war criminals and I'd rather leave any clusterfuck that allows it, whether they do it through will or incompetence.