Perhaps because it isn't a random number generator.

Getting reasonable random numbers with beating clocks,
ala /dev/random in plan9, is not hard.  However, to have
any confidence in it you really have to test it on every
machine.  I've got a few reasonable random number testers
and there are plenty on the web.  I throw them at our
random generator every now and then to make sure nothing
broke.  If in Windows, a process can get interrupted by
the 8253 clock, we can generate similarly random numbers
there though perhaps at a slower rate.  That does require
that the 8253 stay independent of the cpu clock.

However, something like the X9.17 algorithm, seeded by a
small amount of random material will produce a ready supply
of unpredicatble and incoherent bytes for things like
prime number generation.  You don't have to look around too
much, we've already got it.