Crossover distortion has become one of those audiophile myths. It's not a concern in modern
class AB implementations and I have not seen any evidence it has ever been audible, even in old designs which still had relevant amounts. Distortion is easily measured: The Audalytic is a good amp as far as I can see. But simply comparing THD+N specs shows that while it performs well at -106 dB (32 Ω, 1 W), comparable class AB amps like the
Atom Amp 2 will easily best it with below -110 dB under the same conditions. So clearly, a competent designer will produce an equal or better performing class AB amp which is also more efficient (and cheaper) than the class A one.
They are less efficient, but that notion of "most linear" is just an idealized goal with no practical advantage. You can use non-linear components to generate a linear response in a device if designed correctly and it's done all the time. Pretty much all electronic components are non-linear in some fraction of their operating range: Just look at capacitor charge curves or diode forward currents.