Stereophile picture is about right: 1% of about 50V switched at 400kHz. I’m not sure why it bothers you - you cannot possibly hear 400kHz. Speaker will respond to average value. There is distinctive possibility of tweeter modulating this wave with other frequencies, but this would be true only if tweeter’s membrane could move at 400kHz - not likely. Also any effective radiation from speaker cable would require about 1/10 wave antenna that is 750m/10=75m. I wouldn’t worry about that with few meter cables. There is still possibility of cable to cable capacitive coupling but, because of shielding it would me minuscule amount further filtered by amp’s input filters. There is some phase shift at 20kHz (about 20%), but it is usually the case of any amplifier bandwidth limited to 60kHz (my Rowland 102). Improvement in Mosfets speed will allows better initial linearity that is right now pretty good to start with. Class D amplifiers suffer less delay (being one stage only) than multistage class AB amps reducing effects of TIM (resulting in softer sound). Output impedance is also low by design, even without feedback.
Second generation of class D amps was better but I’m not even sure why. Bel Canto’s Ref1000 got additional power supply caps and inductors. That possibly reduced noise on the power cable (Power Factor Correction) but Icepower module used in both generations of this amp has line and load regulated power supply (uncommon in class A or AB amps). My Rowland 102 is plugged into Power correcting Furman conditioner (big cap and inductor plus filters). Second generation of class D amps served one important task - it allowed some critics to remove foot from their mouth.
If top designers like Jeff Rowland, who is very sensitive to noise issues, don’t see any problems in class D or SMPS (that he uses in class AB amps), then I wouldn’t worry about it either and would judge it only on the merit of sound - it is the only thing that counts. If it sounds good then it is good.