Perhaps I have missed something, but what exactly is the goal here?

Some of this stuff sounds like people are trying to make workstations
into atomic clocks or something?

If the goal is simply to have a machine keep reasonably decent time,
having date(1) make an adjustment when you set the time can go a
long way.

If there is a need for a truely accurate time, I would expect to
use some external clock with a special device driver.

If this is a let's see if we can do it just for fun exercise, that's fine,
but put the mods inside #ifdef.  We don't want to lower efficiency for
uucp/slip/ppp/etc or add kernel bloat for people not using the feature.

-Dave