Moving it around didn't seem to matter, but rotating it does make it more or less intense, so I think it's geomagnetic interference. There was nothing magnetic nearby the location that I can see.
Degaussing it in fact didn't improve things, the improvement I noticed before was just that the discolouration fades a bit over time. It never goes away, but it reduces to maybe 2/3 the original brightness.
I ended up deciding to try cheat magnets today. I bought some weak fridge magnets and did some experimenting, and I did find a location that seemed to fix the discolouration without really affecting geometry. (there was already a very slight geometry problem in that corner, but not bad enough to care, and it's no worse) Suspiciously, there are already four small disc magnets on the tube, equally spaced, one in each corner. The location that putting additional magnets "fixed" the problem happened to be... right on top of one of the existing magnets. If this monitor has never been serviced before, these magnets may have been factory installed. The fact that they're equally spaced (one per corner) makes me think that.
Perhaps whatever correction the magnets are meant to make, that one particular corner's thing had grown more severe? Or perhaps that magnet got demagnetized? There may have been a slight reduction in the quality of convergence in that corner, but it's hard to tell, and if so, that is a MUCH less noticeable problem than the discolouration.
In terms of the fix, I put six fridge magnets stacked in that location: they're very weak magnets, and it seems that a bunch are required to fully correct it. Or rather, mostly: when I take a photo, I can see there is still a hint of blue, but it's a massive reduction, so you can't see it with the naked eye in-game.
I took some photos to illustrate.
Here is the unaffected side of the monitor. Note the two disk magnets, and the two white strips of some kind? There are a matching set of all that on the other side:
Here is the other side of the monitor, where I stuck the additional magnets. I used gaffers tape instead of duct tape as a temporary thing, I'll replace it with duct tape when I want to make it permanent:
And here is a shot from another angle, note that I've placed my magnets right next to the existing one (the existing one is on top of the orange double sided foam):
Here is a shot of the red test screen (this is the 240p test suite on an early model SNES), as you can see the blue discolouration seen in my previous post is almost gone. It's worth noting that I moved/rotated the monitor to the other side of my desk to open it up, and the discolouration was actually worse there. Hopefully it remains fixed when I put the monitor back in the original location:
And here is a shot of the test grid, the geometry in that corner was already a little off, and seems unchanged. There may be a slight worsening of convergence, but it's not a huge difference I think. I had previously adjusted the settings of the monitor such that there was only the slightest hint of overscan, although there is several pixels of overscan on the left side of the screen only, because the on-screen menus all disappeared when I had the image perfectly centered. It only re-appeared by either shifting the image two pixels to the left, or by changing the parallel setting by two stops:
Here is a shot from LttP demonstrating the convergence, note that the white borders of the hearts are less sharp in the corner of the screen:
I think it's a lot less noticeable than the red in the image shifting to blue in that corner.