Of course it's audible in stereo-mixed music. You try to isolate things with mono test tones and you wonder why you can't capture it. Sad, weird, to be a scientist with such limited tools.
Rent a recording studio, they need the business, and do some 24bit audio tests there. You will easily hear it in the pan steps, in the depth and clarity of reverb decay, in the presentation of the soundstage as a clear, 3D, thing.
Good signal chain will reveal much, even at 16/44, that most people don't even hear. Even on cheap speakers.
At 24/44 and higher sampling rates, there is more data which provides a more nuances rendering of the music. You will never capture this in the lab.
Science must adapt and grow to learn how to perform proper music sound quality tests.
If you hear 24bit music played on a competent system and you come away thinking that people don't need to hear that, you are anti-music in my book. You can be pro-music and pro-science, ya know.