> Why shouldn't we be able to schedule a timer at sub-millisecond resolution

it's not just the API, the whole operating system need to be aware of the need to operate at such frequency, because even if the clocks are faster, the software gets interrupted all the time. what good would a nanosecond api do if an interrupt can stall the cpu for a microsecond?

hence the need for a whole class of realtime operating systems. the next question would then be, why aren't all computers using a realtime operating system? because with the relaxed time constraints we get great returns in speed and cost.