Courgette meets a dangerous (Red) Bend

But— you protest— that well executed digitization with sufficient bit-depth and sampling captures the information with precision exceeding the noise floor and band-pass of human hearing, and that under carefully conducted double-blind testing even the best listeners can not discern a difference.

Quite right.

But the same is true of lossy compression: At high enough rates with well well enough done methods it exceeds the limits of human perception and produces results which are not ABX-able. It's often not well done, it often is used at fairly low rates, and because it uses more sophisticated techniques it can fail in subtler ways...

This is relevant because your model of lossless=perfect, lossy=bad brings about unreasonable conclusions. Which would produce a more accurate experience: Stereo lossless audio or surround sound at the same (high) bitrate using lossy compression?

Pulling this back on topic, perceptually weighed quantization *is* obvious to a practitioner in the art, or at least it has a very long history of incremental development stemming back to the early vocoder speech crypto devices from the WWII era, the weighing filters used on analog telephone lines and for analog noise shaping (dolby a).

Like in many other areas the underlying technology needed for MP3 existed for a long time before computers became so stupidly fast that what would have seemed like a joke (executing 152 288point complex/complex FFT's per second for the MDCTs in MP3) became completely reasonable. The same is true for some 'recent' innovation in asymptotically optimal error correcting codes.