Cool.
About the consensus, to decide what behaviours are bad behaviour is going to be quite hard, tricky and would take a lot of our time, that's why I'm suggesting each case should be treated on it's merits, if a person sees something he considers bad behaviour and would like to downvote it the rule should state that he brings it up for discussion to the whole Community and the community decides on it.
I'm also saying you should finalize a rule based on this to take care of the '[2] behavior determined to be improper via an approved community-consensus protocol'
Your author's metrics idea is also really great, though, I'd say we finish these malicious downvote talks and then you propose that next.