The solution may be simpler than you think. After months and months of trying to fix basically the exact problem, all it took was a 1 cent cap. Plus some jacks and a die cast case.
The issue I had was with this setup:
PC + focusrite 2i2 interface --->2 balanced lines ---> MOTU interface + Macbook Pro.
The noise I had was exactly as described here: whenever the PC was utilising the CPU or mouse, lots of noise would be present on the MOTU outputs. However, all that was required to cause this was the grounds of the 2 interfaces being connected (ie conductive connection, important to note). The signal lines could be muted or even disconnected and the noise would still be present.
I tried a USB isolator. Total waste of time and money. Not only did the PC not recognise the device, the isolated DC/DC supply caused it's own noise on the powered device.
I tried (cheap) isolation transformers, whilst disconnecting the ground. This resulted in a very loud distorted 50Hz signal being superimposed on the audio signal. Tried reconnecting the grounds, and THAT sound went away, but the original noise came back. Not only that but the transformers had low frequency attention, distortion, and phase shift etc. They were cheap so that was expected.
On a hunch, I tried connecting the grounds capacitively. Well what do you know, it fixed the issue immediately. Turns out there was a constant 78mV DC offset between the grounds. The required capacitance was somewhere around 10nF. With a 2.2nF cap, the 50Hz distorted sound was still there, but sounded different, then with 10nF it disappeared completely. Its odd, it wasn't like attenuation, there was a critical cap value where it just disappeared.
So I tried connecting the balanced lines conductively, but connecting the grounds capacitively with a 100nF PP cap. Issue fixed, no need for isolation transformers.