Why? I think it is very plausible that a large part of this is automated. Combine that with a real name policy and false positives sound not that unlikely. One

Why? I think it is very plausible that a large part of this is automated. Combine that with a real name policy and false positives sound not that unlikely. One potentially error-prone part of moderating any site with community content is handling users evading bans by creating new accounts. If you don't have any mechanism to prevent that, you have a much larger amount of offensive or troll posts, and react slower because you don't have the history of the user because you treat each account as a different person. But any mechanism trying to detect if a new user is potentially a sock puppet of a troublemaker is acting on rather few bits of information. So false positives are not that unlikely.