Your orders of magnitude comparison is off by orders of magnitude.
Yes, you're right: in the supermarket analogy I was introducing 5-6 additional decimal digits (8 instead of 2-3), while in the case of audio we are talking about doubling the sampling frequency and 8 more bits per sample, so what used to be one 16 bit sample become two 24 bit samples, so the accuracy improves by 32 bits or about 10-11 decimal digits. So it would be a more accurate analogy if the supermarket would express its prices with 13 decimal digits, e.g. on the Google playstore you would probably pay $0.9999999999999 instead of just $0.99.
If you're talking about 192kHz sampling frequency with 24 bit samples, you go from one 16 bit sample to four 24 bit samples, so 80 extra bits of accuracy or about 24 extra decimal digits. So on the Google playstore you would get prices like $0.99999999999999999999999999.
No amount of sufficiently high spectral information is going to make any curmulative impact in lower registers (harmonic noise), whereas whole numbers with the same entity will.
The ear isnt a float vector. It interprets a fixed A-weighed range. 24bit is 8 extra bit.
While most of it is outside of the auditory range for humans, its not useless. Also, sub 20Hz information travels in the physical realm, which is a point to make if you arent daft.
Following the obvious rebuttal there, I think having different sound-files for different purposes is a way of the old "lossy for mobile" type ordeal, which got old fast. Being able to do dumb things isnt a reason to do them.
Similar ability in jpegs is just as bad a reason to encode files for other purposes and forget about intent.
Outright bad encoding is very much present within the 16/44.1khz domain of CDs and the like.
To an extent i would rather listen to digitally mastered vinyl, or even casette (!) with what goes on there. The hires _files_ are most often good and done right in comparison.
Wherever you consider the flaw, be it capability, practice, cost or design, its not like there is a lot to lose. CD music is worsening and price of storage is going down.
Design right and pay the price and let the others do as they will, wont change a thing anyhow, unless the power of pyra-sales orders of magnitudes its way to becoming a factor in the market.
With 24bit you can do resampled softvolum and still retain enough dynamics for the auditory range to be saturated, that isnt useless.
In any event, no amount of bad math and ill understood science is going to disdain people from paying the extra dollars it takes to be able to play all files without transcoding first. Its even a quite significant extra value added. (in sales)
