[newaccount74]

Clocks are out of sync all the time, so any code that relies on precise timestamps needs some way to account for out-of-sync clocks anyway.

So if you store two timestamps, you can't rely that the actual time that has passed between storing the two timestamps is equal to their difference, since the clock might have been slow, fast, or might have been synchronised with another clock between the recording of the two timestamps.

I think dealing with leap seconds should work in a similar way as working with other clock synchronisation issues, so I really don't see the point of adding new types of clocks in unix.