Devices with metal enclosure do not approximate faraday cages–remember, bluetooth, wifi, and air are passing in and out–at least. Computers are designed to be permeable.
In my experience, RFI produces the largest effect size of all of the “beyond bit perfect” interference mechanisms. It’s the first place I try to debug when things don’t sound right. And nearly 100% of my bad experiences with RFI have occurred in the presence of plenty of metal enclosures.
When I want to do critical listening for the purpose of evaluating gear, I take the absolute minimum amount of stuff into the middle of a room with no other equipment, and do the testing there with a pair of headphones. Ideally, just a power cable and ethernet cable connect that stuff to a Roon system in a closet far away. I don’t trust what I hear in my office–because there is so much equipment in here and I can’t hope to meaningfully control for who is radiating what.
I’ve reviewed dozens of devices since we launched Roon in 2015. I’ve seen some that like to amplify WiFi signals, some that like to amplify GPU activity, some that like to amplify RFI from the driver circuitry in my Thunderbolt monitor, and some that pick up cell phone signals. Usually when I notice this stuff, the effect size is huge–it’s not something where any would doubt the audibility. Remember back in the 90s when every amplifier would pick up and amplify GSM signals? I’ve run into some products that still do.
Of course, if a device is able to amplify interference to the point where it is unmistakably audible, it’s also amplifying lower-amplitude interference and causing more subtle problems at other times.
I want to stress-this is not just irresponsible manufacturers, and it is not just inexpensive products. Some devices that I like quite a lot, and would recommend to my friends have exhibited this kind of behavior.
There is just no substitute for physical space between the sound critical stuff and the stuff that isn’t engineered that way. The game of “making things radiate less” is moot by comparison. Yeah, you can do it to a point, and see some results, but it’s a much less interesting game.
Anyways–my one recommendation is to look into the system-level optimizations that Audirvana performs. Turning off Spotlight indexing, stuff like that–Roon doesn’t do that stuff, but you can definitely do it by hand if you want and see some of the same benefits.
I don’t think you’ll find that their communication with CoreAudio or their file handling is the source of the differences. If I had to bet–it’s the GPU/CPU load associated with Roon’s rich UI that makes the biggest difference. Not a whole lot we can do there within one machine.