I agree with your first two points, particularly the first, but the third seems wrong.

Third, my expected lifespan assuming cryonics works is not infinite. Say there's always a finite fixed probability k greater than zero that I will die on any given day. Then the expected lifespan I will have is finite (working it out is a nice little exercise). And it isn't like there aren't lots of obvious things that could cut off a civilization even if it had the tech level to do nanotech revival or direct uploads. Even with those technologies, a nearby gamma ray burst could still fry everything.

In general any assumption of a constant per-period risk ensures doom, but we have uncertainty about whether such things exist, and can conceive of scenarios where the chance of destruction per period declines fast enough to give infinite expected lifespan. Any probability assigned to those scenarios gives infinite expected lifespan. Of course, by the same reasoning you have infinite expected lifespan even without signing up for cryonics.