Plain distortion figure (especially THD) is not necessarily very good figure of merit (one might say it is meaningless unless measuring bandwidth is also specified), unless the spectrum of distortion components is also known. Higher harmonics tend to be more annoying even with lower percentage figures. I tend to think that IM distortion is better figure of merit. Most manufacturers seem to just announce the THD, as it gives lower figures. Even if tube amplifiers have relatively high THD figures, the distortion components tend to be low order. On semiconductors, it tends to be other way around.
This is because open loop gain rolls off relatively early for most solid state amplifiers. It really is this excess gain which is used to reduce the distortion. To have really low distortion amplifier, the open loop gain should stay high up to the bandwidth of the input signal. Just look at the open loop gain figures of just about any modern opamp, like
OPA2134. You'll see that the open loop gain starts to roll-off as early as 10 Hz! The difference between closed-loop and open loop gains is the factor which reduces distortion. As it rolls off, higher order harmonics are pronounced.
Then there are also exotic distortions, like DIM (characterized and documented by famous Finnish audio researcher in 70's, Matti Otala). It is claimed that DIM can be heard even if it is below noise floor due to ear function (it can dig signal out of noise due to processing gain).
Here is a paper where he describes a measurement method for DIM, and shows some results (TIM was the term then).
Regards,
Janne