Hi and welcome.
Firstly I suspect it may be hundreds rather than millions, there are not
that many hifi enthusiasts who like doing that sort of test IME!
As far as I know there has yet to be a single controlled comparison published anywhere in the world showing what you assert.
There have been lots of friendly "bake offs" where friends compare things - I have been to several myself, but not in any sort of rigorous way and with mixed results IME.
The main problem is usually time from one "listen" to the next, and the difficulty of matching levels to 0.1dB for many enthusiasts who do not have the equipment required to do it, and often don't realise how easily the results change if this is not done. It is not just a question of swapping components and then listening, which is often all that is done.
One of my aquaintances bought some very expensive silver speaker cables after a demo at his dealer sufficiently amazed by how much better they were to pay thousands for them.
He is an academic and when his nephew asked if he could borrow them for an experiment, effectively some measurements followed by a a blind level matched listening test, as part of his physics studies he was glad to oblige. He went along to the listening test himself firstly expecting everybody to be amazed and confirm his earlier comparison and secondly to discover why.
To his shock and amazement, and believe me he
was shocked, he heard no difference either with a level matched blind comparison although, as he said later, he was 100% sure there was a big gain from them when he first compared.
As far as different components in a device is concerned this is easy to be rational about. The only link between hifi components is between input and output connectors, so if one measures the changes at the output connector whatever is happening inside, be it due to components, pcb layout etc will all be revealed in the measurements.
If we measure the output the signal will have has amplitude, frequency and phase. There is nothing else.
We can measure these to a level better than human hearing, small hum effects of pcb layout or cable dressing, for example, which are too low to be audible but we see on the measurement.
If there is no change in the output of a level that is detectable by human hearing (and distortion audibility level is debated still) then the component
can not actually sound different, any differences heard by a listener are probably placebo effect or non matched levels.
Have you seen this 8 year old level matched blind comparison?
https://www.audio-forums.com/articles/interconnect-cable-blind-listening-test.15/