There is a fundamental flaw in this reasoning, as I see it. Even quantum mechanics puts a limit to the speed of every physical interactions. This means that whatever universe destroying event is happening, it is still constrained to travel at the speed of light. But the universe is very big, and even if the event happens inside our cosmological horizon, it would still takes billion of years to destroy it all.
This violates Tegmark's point n° 2: if the coin flip happens in a region outside the moon orbit, we would have plenty of time to observe it. But since we do not observe them at all, this means either of two things:
1) we are in a very atypical universe;
2) universe destroying events do not happen;
thus undermining the Fragile Universe Hypothesis. To be specific, it's your point n° 2 that doesn't follow through.

But are we in an extremely atypical universe? This is a classical problem with MWI, of which quantum suicide is just a tiny subset: since there's no way to assign a probability to different branches, there's a fortiori no way to assign a probability to those branches that witness extreme violations of classical mechanics (which we do not observe).