I think that will be very interesting if this fft test repeated with a new modified unit, as TomC's unit. TomC repeat this test and with low frequencies like 21.4MHz, please.
I captured some FFT images. The first two are with the scope setup to view GND noise. The next three are with nothing connected to the BNC. In Run mode the signals are random and the capture don't really represent what you are looking for, like spikes at the 250MHz and vicinity of 25MHz.
So what I did is set the trigger to Normal for the GND noise images, and bring the level up until it no longer triggered, then back until it triggered once in a while. This way I could capture when the highest noise spikes were present.
For the images with nothing connected to the BNC there is not enough signal to trigger, so I had to use Single trigger mode and Force trigger until I got some images that showed the area of interest. But with the scope set to 5MHz/Div if you are in Auto trigger all you see is random, so I don't see anything significant in the 25MHz vicinity.
Now that I have a spare PSU, I've been examining it and trying to draw a better schematic than what they have on Version 1 of the Service Manual. The one thing that struck me as odd, so far, is that there isn't a bypass capacitor anywhere between either PSU ground plane and Frame ground. However, there is one between the two ground planes. I haven't checked to see if that's something that rf-loop added on his DIY version 2b. Let me know what you think.