Analogue, and it not being digital... and why 192k could be useful, not audible
Bravo Stoneshop for mentioning the -3 dB measure. A lot of people talk about "Bandwidth" of analogue systems and think that if a system has "20 kHz", then nothing beyond that is registered when that's only true of digital sampling. In practice, the recorded signal level rolls off beyond this point (just as human hearing does).
I'm surprised that the author didn't consider the anti-alias pre-filtering of the original analogue material when producing CDs. Often the fitering starts at around 16kHz with gentle rolloff, rather than 20 with a sharp cutoff, because CD's high-frequency phase response isn't particularly good (a problem that's solved by using a higher sampling rate, although beyond 96 kHz, it's hard to make a case on "listener" grounds). Also, sharp-cutoff filters have "leaky" passbands, allowing alias signals through: another problem fixed by using higher sampling rates.
I accept that nobody can hear anything much over 18 kHz, and most adults with the money to spend on fancy hi-fi can't even hear 15 kHz anymore. That is statistically speaking, however, and there are a very small number of people who retain a wide hearing range well into their 40s and 50s, but never much beyond 15kHz. But it's still odd to see a defence of the 44.1kHz sample rate of CD as if it's scientifically based: it's not, it was the highest rate that could be placed on NTSC U-Matic video-tape without the machine's horizontal-blank insertion destroying some sample packets. The fact that every subsequent format has used 48kHz or multiples of it, shows that it wasn't quite good enough. 16 bits also wasn't quite enough for preserving linearity of low-level signals; 24 appears to be the minimum for preserving everything that's audible as the signal goes through its various stages from studio, to medium, to replay, to ear. (I'd argue that the extra bits of the "high res" formats are more significant than the extra samples)
There are also a couple of non-technical advantages of higher bitrate formats: mainly, they require components and system makers to work to higher standards. Before CD arrived, most amps (non-hifi) had a 15 kHz top end, because that was good enough. As noted above, most adults can't hear beyond this anyway, but an amp designed to only reproduce 15kHz is going to be working reasonably close to its fastest swing when reproducing a very audible 8kHz, and failure to track signals accurately will increase harmonic distortion. Signal-to-noise was also not a real issue at this time, as all sources had quite high noise floors anyway.
Suddenly, CD players arrive, and amplification needs to nominally support 22 kHz. That means new ICs and transistors (valves, if you're rich enough to run Class A, and not care about your electric bills) with a faster slew rate (the speed at which an amp can change output voltage), which means more accurate tracking of signals in the audible band. Similarly, the lower noise level of CDs pushed equipment makers to produce lower noise amplification, which is a good thing in general, because for other, analogue, sources the noise floor still had small amounts of signal buried in it; signal that a lower-noise amplifier would now not obscure.
So, if 24bit, 192kHz audio became commonplace, we'd have amps that are good for 30+ kHz, with an improvement in linearity in the audible range and noise. What's not to like? So what if it's driven by marketing fluff? So was CD, and it raised the bar.
But the real, hidden benefit of 24bit 192kHz is something that the author almost touched on but then left hanging in the air: and again it has nothing to do with human hearing. Room-tailored reproduction. If you're designing an active, adaptive speaker system to compensate for the room's acoustics, your task is made a lot easier if you have a source signal with lots of resolution, as rounding and errors are less significant, and lots of samples, giving finer control of signal phase..
Right now, home cinema system DSPs already upsample to higher rates before applying processing, but upsampling a signal is no match for working with more original data. It doesn't matter that this extra data is inaudible to listeners; it still allows the processing equipment to produce a better audible signal after applying lots of other processes to it... after processing, a CD-quality signal can never have the fidelity of the original (you cannot add information by processing a signal, only remove it). On the other hand, a 192k/24 signal, after processing, can be every bit as good as the unprocessed theoretical CD signal.