Well thought out, and I like the idea; especially how it could potentially remove the negative connotation of flagging due to disagreement with reward.

My only concern would be the implementation as a custom_json_operation. Granted, we can do it now as just a Steemit thing and not a steem thing, but I fear that we're going to have a large amount of bloat due to adding so many TX's. Though, as most votes are likely via bots and not from the Steemit UI, it might not be so bad.