I don't have experience in active noise cancelling design, but I would guess that lower latency hardware probably plays a large part in the improvement, given the nature of the problem - a chip set that couldn't 'keep up' with changes in the audio signal would be late in responding, resulting in comb filtering. No doubt, people also improved the algorithms over time too.
Ok, so the design improvements made sounds more synced, but imho that doesn't solve the problem (it might even make it worse? Not sure...). The combing effect is there because the sources of the sound and the "anti-sound" are not in the same position. Lower latency helps that the signals are perfectly cancelled in some predefined direction, but you still get twice the sound amplitudes at a different angle (towards the head).
I don't have experience in active noise cancelling design, but I would guess that lower latency hardware probably plays a large part in the improvement, given the nature of the problem - a chip set that couldn't 'keep up' with changes in the audio signal would be late in responding, resulting in comb filtering. No doubt, people also improved the algorithms over time too.