The following solution came to me a while ago, on a different issue about bots and how people (mis)use them.
If we could indicate that a user is a bot (similar to flagging) this could be determined through the feedback people give and it could be known in the system. Then there could be different rules for bots, for example no comment limit, no curation rewards etc. (few humans would need to comment more than every 20 seconds, unless they have serious hyperactivity or something)