But it also has to be defended separately by the admin of every server that has a user subbed to that community. Seems like a large burden to put on small-mid instance admins.

I’d be surprised if my server admin was really paying attention that closely to votes on communities I’m subbed to, right?
I have to admit I don’t know the view that admins get of how their server intersects the fediverse.

I’m not sure how giving every server access to the votes solves that.
The malicious server can make fake users to pump up votes. your server admin has to notice, then check the vote logs, then see what’s happening and defederate them. That’s pretty much what you described in your scenario, anyways.

What do you mean “send fake votes”?
Or rather, who do you think should be responsible for identifying and blocking fraudulent votes?

And how do you reconcile votes that come from servers that you’ve defederated with? Should everyone have the same view of the post, or should people only see votes from servers that their server is federated with? What about votes from users you’ve personally blocked? Etc

I personally kinda think that the responsibility is on the server hosting the post, and that everyone should see the same (but anonymous) vote count, of which the hosting server is the single source of truth.

The server hosting the post needs it.

It only needs to tell other servers the vote count, and the votes of people on that other server.
That may not be how it actually works, but that’s all that’s needed