Point is, if Alice and Charlie both behave exactly the same, both notice Bob does something bad, like calling in bid bots on a post tagged for one of the top curation initiatives (something I've used flags for myself a couple of times, and have been retaliated against a couple of times as well), and Alice was the only one acting on it by attenuating the bot votes with a down vote, then behaving like Charlie is incentified above Behaving like Alice by the retaliation by Bob.
For the flagging to truly work without negative incentives, I think we would need some kind of flagging SP that can be delegated separately from regular SP to semi-automated services. Without something that can shield those who for example want to use their free downvotes for attenuating false curation bot upvotes against direct retaliation. As the EIP comes with nothing that shields us from retribution down votes, I'm afraid the incentive to act is still negative.
In this case A, B, and C did not behave exactly the same.
If you think that literally everyone will downvote and/or get downvoted and/or retaliate and/or get retaliated against all in lockstep or that literally no one will and we will all drown in a sea of sameness, we will have to wait and see but I'm very much doubting it.
Somewhat. Certainly there is still minimal if any positive incentive to downvote, and indeed many still may just not bother, or not want to enter the fray. That is a clear potential point of failure. However, all those issues already exist. Compared with now, the overall cost of downvoting is surely reduced dramatically for nearly everyone. How much difference that will make remains to be seen.