Coupla' comments, bearing in mind I am no expert on audio amps:
1. Many amplifiers exhibit low-level distortion at low power but they are not typically tested there. Looking through a few old Stereophile and Audio test results, crossover distortion is fairly obvious at 1 W when they do they test at low power (rare). I tend to believe (but do not know, no expert) this may be exacerbated by the trend for multiple amplifiers in one chassis and subsequently lower bias current. Crossover distortion is pretty ugly; I am not sure I would call it IMD though that is a by-product. It can generate a series of spurs based on the waveform that has both harmonnic and non-harmonc content. For a data converter designer, it is akin to sampling glitches from a T/H or at the output of a DAC.
I am going to have to rethink my assumption that the reduction in SINAD (THD+N) at low power levels is usually noise-dominated. It looks like that may not be true these days.
I don't know if the hysteresis of a class D amp contributes in the same way as crossover distortion; IME the error waveform (looking at it extracted from the output with the fundamental notched out) is somewhat similar but frequency distribution quite different and moved above the audio band where it is more innocuous. Of course class D has other challenges... As does any other class.
2. Last I checked the THX spec min load was 3.2 ohms, though my memory may be faulty. The trend for a number of years now has been lower impedances and there are quite a few speakers besides ESLs (which often go below 2 ohms) that are in the 2-3 ohm range. Heck, even my old Maggies measured something like 2.7 ohms at HF where the ribbon took over.
FWIWFM - Don
Edit: Crossed posts with John, who is an expert, and at least my comments don't appear completely out-of-touch.