let me step in here. Audiotroy, you are wrong on many fronts and don't understand how hardware/software works.
"electrictrical noise"? You bring up cpu's when you tried to explain noise. The cpu's that any 3rd party server uses are not specifically built to run Roon. For example, Auralic runs/used to use the Tesla G1 cpu. Do you think the Auralic engineers went to the Tesla chip manufacturer and told them to make a unique processor to run Roon? Do you think Aurender engineers went to Intel to ask them to build the chip they use to specifically run Roon? Come on, nobody believes that, and Intel/Apple/AMD would laugh at somebody that would come in to ask them to build a specific chip to run a specific app that you would only sell a few thousand of them, if Intel/Tesla would design/build such a cpu, it would cost millions of dollars for each cpu. Aurender claims they run a low powered Intel chip and you stated the same, but if you compare Intel cpus to Apple or AMD Ryzen cpus, intel is not low powered, why do you think Apple went away from Intel.
Another misconception you have is your statement: "We run Roon in its own CPU core". You have no control over this, it's how Roon is built, its out of your hands. Roon uses multithreading which utilizes a single cpu, it isn't written to multitask/parallel process over multiple cpu's. This is why the bigger the music library is/ or the more endpoints you have, Roon will eventually saturate the cpu its running on, that's why if you have a large music library, you will need a faster cpu. Most enterprise software multitasks/parallel processes over all the cpu's in the server to prevent such bottlenecks.If you look at your system resources running Roon, you will notice that cpu is getting saturated whereas the other cores are idle.
Your years of testing are flawed IMO. What would you test? What does your testing actually consist of? You bring up cpu's, do you actually test the noise difference between an M2 Apple cpu vs a Tesla G1 cpu vs the low powered Intel cpu? No. Do you actually benchmark different flavors of linux? No. Every streamer is a server, it has a cpu, memory, and an OS. How many iterations of each of these components have you tested? Did you test the Tesla G1 cpu using Ubuntu v18 vs Ubuntu v20? How about Fedora?
Bottom line, you don't know which cpu is the noisiest. You don't know which OS is the most efficient or the best OS for say USB or i2s. I do know that most of the streaming servers run a proprietary/tainted version of Linux, and with this, you either have to have the manufacturer update/or fix any OS issues, or if they are out of business, you will have to hire a Linux hacker to tweak the system. I don't use any hacked/proprietary/tainted Linux on any of my servers.