@srsa_4c
This is excellent. Everything now is back to normal operation (for the last couple of hours anyway without venturing too far elsewhere). CWB continues to work OK too. I am finding that novigncorr (after a CWB) makes the cam super-sensitive to the illumination setup and CCD window cleanliness. To really go down to the few units of intensity consistency, I've had to change the Teflon LED drape to a flat panel sandwiched between two microscope slides. Despite improved source uniformity, I feel assured now that the camera is telling me my test setup is still very slightly deficient in achieving perfect source uniformity (by about 3/256 intensity units), favoring one side of the CCD. After taking many test pairs slightly space shifted (one pair here for viewing) I could see that slight depression moved around and I thus feel confident that there are no more lens distortion corrections left.
Now here's the very nice *additional* news (if what we're getting so far isn't good enough already): the centers of both images match intensity exactly, and in the correction (Fig 1) the corners are gained up relative to the central zone. That doubly-confirms Ev performance is maintained through your hack! The upshot too now is that this is the only S90 on the planet (for the moment) that has a whopping 30% more dynamic range in its RAW formats, uniformly across the entire frame (Fig 2)!
I still have more tests to do.
edit: Our "instrumentation camera" told me to improve the projection setup a 3rd time by removing the collimating lens and Teflon, and instead creating a distant point source by moving the LED to 1m away from the CCD surface, installing a shroud (black-painted toilet paper tube) and then doing the captures in total darkness at night. These figures are the result, and that is the performance we need. If you mouse-over pixel-read Fig 2, the whole system works perfectly, +/- 1 count over the entire frame: you could think it's a fake!
CCD @ +18C, DFS, 1/8s, ISO 100