I don't find it hard to believe that very low levels of certain kinds of distortion are detectable when at the same time virtually un-measurable differences between passive elements like cables and capacitors are widely believed to be audible.
THD-20 does not tell the whole story on amplifier behavior, but it has also been given a bad rap because single-number usage of THD-20 can obscure the real story. If you have THD-20 with a full-spectral analysis out to the 7th or 9th harmonic, it tells a lot more, but most THD analyzers start cutting off at 80 kHz. One should always take note of the waveshape of the THD-20 distortion residual as well. If it is rather sinusoidal, that is often not too bad. If it is spikey, that is potentially very bad.
CCIF 19+20KHz with full spectral analysis is much better because the IM products lie in-band.
At the same time, moderate levels of benign THD, like mostly 2nd and 3rd, may not bother anyone very much. This...