If the only votes that would really matter are from the big stakeholders, is this part even necessary?

Strictly speaking, I guess not, but I think there are three reasons why you'd want them to. (i.) skin in the game. The need to expend voting power and resource credits would impose some restraint on the signaling; (ii) the moderation node could track whether participants actually followed through with their votes and ignore signaling from accounts with unreliable follow-through; and (iii) ideally, at some time in the future, the program could have an influence without backing from a top-tier "sponsor". Also, if this were adapted to fund itself using upvotes, then the participants would need to vote to collect curation rewards.

I was trying to think if there was a way to avoid this element, but I'm not sure there is.

Yeah, I don't see any way around it.

But might it not be the case that there would be few enough genuine humans participating that a vengeance-minded person could just mass-retaliate against all of them even if they couldn't find the particular person who blew the whistle?

It might, but that just sends the rewards back to the rewards pool for redistribution, so I'm not sure how much damage a broad-based retaliation would actually do. OTOH, they could just randomly scapegoat individual accounts, one at a time, so I guess there would need to be some sort of mitigation plan for dealing with those contingencies.

Yes, the perennial problem.

Right. And the only tool that we have to address it is persuasion.