Not really, though -- it was empirical, and the fact that an estimate of the first halving does X doesn't imply that further halving will have the same effect,

Not really, though -- it was empirical, and the fact that an estimate of the first halving does X doesn't imply that further halving will have the same effect, so a reductio doesn't work here.