Note that zero crossing detection is easier to do by picking up the voltages on the secondary side and have a semiconductor start to conduct when the voltage (positive or negative) gets far enough from zero.

But anyway - you are still thinking about taking one measure every half period. That only works if you have a normalized value to pick up at that single point in time.

My discussion about residual does not intend you to take a single measure of input voltage and compare with a single measure of a reference sine wave. That would still not tell you your "energy error". It would say that there is a voltage error at that specific time but will not tell you if you have too high/low voltage or if the difference is caused by distortion. Remember that with distortion, the average may be correct (expected RMS) but the difference between sine and reference will ripple because the distorted signal contains overtone frequencies.

If you do look at a difference signal, that signal (if the amplitide differs) will change sign everh half period. And you can't rectify the signal since you need to known the sign of the error to know which of the two signals that is stronger. In some situations, it might be possible to rectify/filter the signal to measure an analog signal, while also using a comparator to get a digital signal telling if the op-amp output was positive or negative. Matching the comparator-produced digital signal with the current half-period would tell if you are too high or too low, while checking the analog measurement would give an indication how much off you are.

One issue here is of course if your device should continue to feed power if you get too much distortion, or if too much distortion is a reason to cut the power to avoid an overload situation.