I've just had a pencil and paper out and sketched up what happens when you stick a piece of glass between the the lens (or mirror) of a telescope and the imaging sensor. The halo is (as I noted earlier) 2 * glass thickness / (focal ratio * glass refractive index * pixel size) in pixels. If the star is in the centre of the chip, then the halo will be centred on the star. Away from the centre, though, the centre of the halo will be offset from the centre of the star by an amount equal to 2 * thickness of the glass * distance from the centre of the chip / (focal length of the scope * refractive index of the glass * pixel size), again in pixels. So as long as all of the halos are offset away from the centre of the frame, that is normal and does not indicate any tilt in the system. Any asymmetry in the appearance of the halos may well indicate that there is a tilt in the imaging chip relative to the focal plane, but then that would also show up in the appearance of the stars and how in / out of focus they were, so you could pick that up more accurately with a Bahtinov mask, for example. An artificial star would be pretty good for doing that kind of work. Interestingly, by my calculations, the distance between the piece of glass (filter, whatever) and the imaging chip has no bearing on the size or offset of the halos, so if distance does have an impact, then there must be something else at play.
I've been focussing my efforts on the filter as the source of any internal reflections, but of course the AR window (or IR cut window) is also a possible source and these things are 2 mm thick on most of the ZWO cameras, in other words about as thick as a 2" filter. I guess that things are just much worse with narrowband imaging due to the ability of the filter to really darken the background allowing the halo to shine through in all its glory :-(
If halos are being caused by reflections between surfaces on different elements, for instance between the camera window and the filter, or camera window and the final element of the lens system (I think that is unlikely in the RedCat because that final surface is curved as far as I recall), then the formula for the halo size is a little different at 2 * separation / (focal ratio * pixel size) in pixels. So, say you had 55 mm between the filter and the camera window (I use that because of the usual 55mm back-spacing), any reflections between them on a RedCat with an ASI2600, the halo would be 6000 pixels in diameter, or so bigger than the sensor, so not visible. But a 2 mm glass window / filter would create a halo of 145 pixel diameter.
I have tried and tried to remove halos in post-processing, but I've never had much success - maybe I'm too impatient. I clearly need to change my perspective and start to try to see them as a feature of a shot rather than an imperfection!