Also decided to use ASTAP's built in CCD inspector (hey it's free).
First image is from about 5 frames into the sequence, showing "moderate" tilt 0f 19%. That's off by a little bit compared to CCD Inspector (which showed 22% for the same image) but the trend is the same.
Comparing this to the image post-meridian flip again, which was down to 12% (mild tilt).
I think the fact that the meridian flip seems to have "fixed" the issue, tells me the issue is probably in the imaging train? Maybe something not tightened enough.
Adding to this, there are some definite back-focus issues which I already know needs fixing.. boy this is going to be a fun project to fix