This results in much smaller and more efficient machine code.

It should not affect correctness, either: Indeed, the delay value
is computed from dividing an int by another int, so it fits into an int.
The delayTolerance is a fraction of that int, so it also fits.
Intermediate results like delay * pro.zero.high are ints because they
are equal to
 timings[0] * pro.zero.high / pro.syncFactor.low
but syncfactor.low is always larger (at least in all current protocols)
than any other high/low value.