Ok here goes.
I am a friend and colleague of Max Townshend but nothing to do with the cables just transformers and electronics. This is just my take as an EE.
Characteristic Impedance applies at all freqencies The value can be calculated at a particular freq using Heaviside,s Telegrapher,s equation using the resistance, capacitance , inductance, and conductance per unit length of a cable. At low freq the resistance term dominates until at higher freq the other reactive terms come into play. So a nominal 75R coax is only that value above 100khz. There is an article on the Belden website covering this in more detail.
Why the CI of a cable can matter is because if the cable is not terminated by an impedance which is the same value as the CI some of the energy in the wave being transmitted is refected back down the cable towards the source end.. As the source is usually low impedance there is a similar reflection and so on All this though is at 0.8 the speed of light or thereabouts so shouldnt affect events at audio frequencies, or should it ?
Its certainly easy to see these reflections on a test set up. If you pass a bandwidth limited square wave down a reasonable length of cable which is loaded by a mismatched impedance and view the waveform produced with a good scope you will see the reflections as a series of steps on the leading edge.. There have been some papers produced on this (Delatraz) if anyone is interested in repeating the test. Ok some will say "but this isnt music" which is true but a square wave does contain a wide range of frequencies so I think its valid. This can also be simulated in LTSpice which does show waveform distortion but only at freq above the audio band. These could affect certain amplifiers but its difficult to prove.
One way being developed is to drive identical signals, including music, down two identical cables and load them differently. Then analyse any difference between them. You can actually listen to the difference as well. Of course you could also use two different cables (or lengths) and load them the same.
Its producing some interesting results but it has to be made certain that only the cable differences are being measured not for example slight amplifier differencies.
I think the intention is for the test method to be made public for all to try.
Ps I forgot to explain for a loudspeaker cable to try to match a typical loudspeaker impedance so as to minimise reflections it has to engineered with very low resistance per unit length and certain inductance/capacitance ratio. In this way the cables CI can be near the speakers impedance above 1khz . Paralleling up CAT5 twisted pairs is a good way