I note that for this argument to go through, you need that our utility function over different multiverses really is convex, which you didn't argue for.

I would agree with that (on grounds of excessive duplication of almost-identical agents, for one), but the point for this post is more to get people reactions than to push a particular theory.