Colin Plumb has an amusing quote in the bignum library, where another
algorithm is used that has a failure probability:
"The chances that another Dinosaur Killer asteroid will land today is
about 10^-11 or 2^-36, so it would be better to spend your time worrying
about *that*."
This is assuming that such asteroids hit Earth every billion years or so.
It also assumes that they aren't detectable in advance, which according to
a documentary I saw recently about the late Dr. Gene Shoemaker, is
correct.
Hal