My experience aligns with @DonH56I have a TOTL Marantz PM-95 (and a PM-94 class A/AB) MOSFET integrated from 1990 with switchable Class AB (120wpc) or Class A (30wpc) and apart from the increased physical transformer noise when switching to Class A and the massive increase in heat produced, the subjective difference is zero, that is until you run out of power in Class A and it is very obvious.
A huge amplifier, 27.5kg built on a diecast chassis, all copper plated and several kilogram heatsinks to get rid of the heat in class A.
Guess who sent @amirm a TOTL Marantz PM-90? That's one generation newer than the PM-95, and drops the on-board DAC resulting in better "advertised specs" of 111 dB vs 106 dB SNR and a THD of 0.0015% instead of 0.0080%. It's the last Marantz before HDAM was introduced in the PM-99SE (which resulted in a poorer advertised THD of 0.0050%)?
It's also interesting in that it used dual EL transformers but it is not a dual mono design. The use of two transformers is reportedly designed to allow the flux leakage to partially cancel out.
I have an Accuphase P-266 in addition to the Marantz. Accuphase's own data suggests that Class A decreases the 20 kHz distortion but in the 1st watt, it's actually worse for the 1 kHz and 20 Hz distortion compared to their AB mode. I'm really curious to see the 32-tone test on the PM-90 between Class A and AB. It may be interesting to test the PM-90 at 100mw, 1W, 5W, 10W in Class A vs. AB.