Fair point, and haven’t seen this forum’s thread.
Douglas Self has demonstrated that optimal bias provides the best THD overall. Designs that overbias, allow you to have pure class A performance across the listening range with the consequence of transients having higher gm-doubling. Some low end designs may underbias to save money on heatsinks.
This product is nice in that it shows measurements for Blameless Class B against pure Class A
http://www.signaltransfer.freeuk.com/trimodal.htm
Optimal bias will determine w A transitions into B and you can do the math to get optimal numbers for the lowest overall THD.
I would think Denon/Marantz set their bias for optimal performance, not for class A at low output, well, may be the first 0.05 t 0.1 W at the most. However, I do think they are probably designed/set on the high side of optimal. I could be wrong as it just my hunch, seeing that comparable Yamaha units that are also class AB and have no visible better heat sinks would run notably cooler according to forum talks, that could also be unreliable. Anyway, it just a guessing game.. May be Amir can ask his contacts for comments on this perceived, and/or potential "hot" issue.
By the way, I don't know why there are so much talk about class AB implies a transition to B at some point, my understanding is that a real class AB amp is by definition class AB, that is, with bias optimally chosen so that both output devices in the push-pull cycles would conduct for a small angle of overlap in order to reduce crossover distortion. Wiki's definition seems to be among the clearest that I could find. That overlap angle is always there so again, it does not transit to class B as such. I suppose there could be amps advertised as class AB that actually transit to B in order to be more efficient, and run cooler. If so, I would avoid those amps.
As I cited earlier, if you compare the AVR-X3600H/X4700H's output graphs, you will see that they did very well at the low "first watt" range. Based on what I read from the Passlab website, my educated guess is that it might be an indication of class AB bias, while supposed to be "optimal", it might also be on the high side of optimal, resulting in a little less distortions at low level, but a little warmer even when idling or at very low output level.