I started to explain the math and physics but figured you won't follow it anyway. So instead, I did a quick experiment of your scenario of coax cable with shields not connected to anything.
I put on an RCA jack on a BNC adapter and hooked it up to my Tektronix scope:
View attachment 8381
In this case the shield was connected at one side (scope side). The output is very clean.
Now let's see what happens as I pull out the RCA just enough for the shield to disconnect as you say you are using:
View attachment 8382
Immediately we see large increase in both low frequency coupling (the large sinewave) and high frequency (general fuzziness of the waveform).
Let's do the test again, this time just draping it over the front display of the scope. Again, first with ground connected at the scope:
View attachment 8383
Very clean again. Now let's float the ground as Mike has:
View attachment 8384
Notice the large increase in high frequency components plus some spikes in the middle of the waveform which is deterministic (i.e. some oscillator inside the scope running at that frequency).
In other words, a floating shield is useless when it comes from shield against both low and high frequencies.
Yes, there is such a thing as reflective losses but that works on principles of plane waves, usually far field, etc. The metal enclosure for your audio gear works in that regard. Which would have been another reason to put the switchmode power supply outside.
This experiment took me 3 minutes to conduct. Mike doesn't have any instruments, nor would he know how to test anything with them. So instead, he goes by stuff he reads online and that is the most dangerous thing a novice could do.
These topics are complex. You can't become your own brain surgeon. The rules of thumbs for keeping things short, separate, etc. were created to guide the people who don't understand the topic but still want to produce good products. Mike flaunts all of that due to total lack of any design experience.