Never heard of that - where did you read it, Wikipedia? :-).
"Pretty much" for some purposes. Just as amplifier distortion can be detected better with the "two tone" test than with a single tone, non- linearity of air is more readily detected by mixing two ultrasonic signals than by looking at a single tone. A 140 dBa ultrasonic sound is completely inaudible to a human (and probably causes no damage), and it only needs to produce something like a a 60 dBa audible sound right at the ear to be clearly audible. Normally distortion of an audible signal that far down would be pretty much insignificant, but here it is not. Efficiency not required. These systems modulate both frequency and amplitude of the ultrasonic transducers, since that is what is required in the audio frequency output. No doubt you can find a complete description in some Journal of the Acoustical Soc. Am. article, this is old stuff, first heard it at the '64 worlds fair.
The non-linearity of air is also important in the production of harmonics in wind instruments, where sound levels are quire high inside the instrument, and is the reason shock waves can propagate. (Liquids and solids are non-linear too.)
Speaking of shock waves, the N-wave produced by a spark gap in air can have a rise time well under 100 ns, and is a good way to check the frequency response of a microphone up to a few MHz at least, with a bit of attention that reflections arrive only after the rise time measurement is complete and the shock wave is adequately square to the microphone surface. A centimeter or so spark gap is good, discharging a capacitor of a few uF, microphone a couple of meters away so that the spherical wave looks fairly flat at the microphone. A microphone with suitable frequency response is the only reasonable way to check the output of a piezo tweeter. We built some microphones which were flat to 2 MHz at U of R in the late 60's; not sure what is available off the shelf today.
Glen