I spent this evening taking 'pseudo' flats with the camera just pointing into an evenly lit wall. I tried a couple of different things to pinpoint the cause of the banding:
*Using a brand new 1m USB cable instead of my old 4m cable that runs under the obs floor.
*Running it through a USB hub or not.
*Connecting the QSI power adapter to another power point.
*Connecting the QSI power adapter to a 110v adapter
I created a stack of 20 flats for each configuration. Result: All have exactly the same banding pattern... so it must be a property of the chip I guess. Maybe that's just how ccd chips behave?
The good news is that when I then divided one of my stacked light images with the stacked pseudo flat the banding disappeared completely.
See attached screenshots: First image is a corner of the psudo flat showing the banding pattern, second is light image before flat fielding and finally after applying the flat (Ignore the gradients, that was still because of the full moon...)
So the problem seems to be solved, I just need to take proper flats now and add them to the calibration routine.
Though I'm still curious to know if this banding is something others have seen too? I can't find any mention of it on the web.