As you can see, the NC252MP reaches the maximum theoretical peak efficiency of an AB amplifier (78.5%) at 75ish W into 4 ohms. It's still close to 70% at half that, and still over 60% at 20 W (which is 8% of max output). And it's not too much worse with 2 ohms.
I had a look for a graph for AB amp efficiency vs. output and found this:
By about 8% of maximum output, you're down to little over 20% of efficiency.
There is absolutely no doubt in my mind that Class D amps have much better efficiency than their AB counterparts (as you would expect), with idle power alone being about a factor of 4 or so lower. Making them multiple times more powerful is obviously eating up a good part of that advantage, but does not fundamentally negate it. And there may still be more to be gained on the power supply front at idle. Looking at the above, a pair of modules would seem to be drawing about 10 W on the DC side, and there is no reason why this should have to be more than 15-20 W AC. Are typical Class D power supplies even using resonant LLC topology?
In the past relatively few people would have considered going for a power amplifier that can deliver 200 W into 8 ohms or 250 W into 4 ohms, and you would have been looking at something with at least an 80 W idle power draw. There's a reason why PA amplifiers in particular went Class G decades ago.
If you want a "traditional" amplifier that is competitive with Class D, it needs to be at least Class G, or better Class H altogether - at which point you are basically using a Class D amplifier to drive the supply rails. That brings with it a set of challenges, of course.
Now the fun part - could you apply the same techniques to a Class D?
Maybe? I could certainly see something like a tiered power supply working where the high-power part only kicks in when needed.