Well, things are less clear here. Prior to MQA being rolled out, Bob Stuart and crew published peer-reviewed research/double blind tests showed that people could tell the difference (to P<0.05) that if you resampled the high-res file to 44.1 kHz, the effect could be audible. Therefore, their mission became preserving the high sample rate, not because people could hear ultrasonics, but because filters could impact the audible band.
Bob took this to another level then saying "timing" in audio matters and therefore if you can preserve 192 kHz sampling, you should. Don't ask me to defend this bit because I can't.
There is also some conjugate filtering which in practice doesn't seem to be in use.
My position in all of this is different. If a file is available in 24/96 kHz, I want it before someone tries to convert it to 16/44.1. I have no need for their conversion. That conversion is lossy of course so all this talk about MQA being lossy is for not. My other hope was that high-res audio would come without loudness compression. In some of the AB tests of MQA to no-MQA content, it is clear to me they have access to better masters than what is already released. That, makes an indisputable difference in fidelity.
Indeed, I know of no one who has made it their mission to try to encode high-res content for us as MQA has. They are likely spending some effort to try to find better masters at times if what I heard in demos is true.
Anyway, I get MQA for free in Tidal. Roon decodes it for me for free as well. Someone wants to cry that I am getting ripped off, doesn't have a leg to stand on.