Bostrom's estimate in "Astronomical Waste" is "10^38 human lives [...] lost every century that colonization of our local supercluster is delayed," given various assumptions. Of course, there's reason to be skeptical of such numbers at face value, in view of anthropic considerations, simulation-argument scenarios, etc., but I agree that this consideration probably still matters a lot in the final calculation.

Still, I'm concerned not just with wild-animal suffering on earth but throughout the cosmos. In particular, I fear that post-humans might actually increase the spread of wild-animal suffering through directed panspermia or lab-universe creation or various other means. The point of spreading the meme that wild-animal suffering matters and that "pristine wilderness" is not sacred would largely be to ensure that our post-human descendants place high ethical weight on the suffering that they might create by doing such things. (By comparison, environmental preservationists and physicists today never give a second thought to how many painful experiences are or would be caused by their actions.)

As far as CEV, the set of minds whose volitions are extrapolated clearly does make a difference. The space of ethical positions includes those who care deeply about sorting pebbles into correct heaps, as well as minds whose overriding ethical goal is to create as much suffering as possible. It's not enough to "be smarter" and "more the people we wished we were"; the fundamental beliefs that you start with also matter. Some claim that all human volitions will converge (unlike, say, the volitions of humans and the volitions of suffering-maximizers); I'm curious to see an argument for this.