Bart, I have a similar noise-based way of deciding when to stop turning up the ISO that has nothing to do with unity gain, although my targets are much simpler; just one D65 patch.
I prefer your current presentation over the above quicky that I did some 4 years ago. Despite a larger number of variables (non-uniformity of the CC patches, light/exposure, and the Raw converter) in my test, total noise also approaches a more constant build up at or above unity gain, but it also illustrates an overall increase of the noise level. Your presentation allows to find a sweet-spot (if there is one) in the S/N ratio, which is very useful for Astrophotography, because the overall noise which may be a distraction (and problematic for sharpening) in regular photography can be reduced by averaging multiple exposures quite dramatically.
I've read your blog, and am pleased to see the progress you've made in finding a relatively easy to interpret graphical form to express the complexities underneath. I'm also pleased with the attention given to characterizing/reducing methodological errors with respect to sensor temperature and shutter speed variations.
The only improvement I can currently think of is an elimination (or at least determination) of the influence of non-uniformity of the light-source and lens vignetting, and sensels (PRNU / sensor dust / dead- or hot pixels), in the central crop area used for analysis. The only way would be to do a check with subtracted image pairs (and Stdev/Sqrt(2)), or (slightly less accurate but still informative) a comparison between the 2 Green filtered sensel sub-images. It may reveal that your current results do not vary much from an even more normalized data-set, but dust and PNRU do keep lurking around the corner, waiting to strike.
I find e.g. that there is a (negligible) small systematic difference in the mean
sensel values from the G1 and G2 sensels of my 1Ds3 camera, the G1 sensels on average give an up to 1.605 ADU higher response than the G2 sensels. It's nice to know that this will usually not bias the noise statistics in a noticeable way, but still there is a difference. It also does allow to reveal shortcomings in isolated Green filtered sensels in the area under investigation, which may influence small crop area results (see attachment, with seriously
boosted contrast, which reveals a slightly less responsive G1 sensel).
I first do a series of exposures of the target at various ISOs, using only the central 200x200 pixels, and holding the digital values constant, in this case 5 stops below clipping for the green channel. That means every time I increase the ISO by a stop, I have to stop down a stop (because I'm compensating for the read noise, which varies with shutter speed, I hold the shutter speed constant.)
There is a slight (nitpicking) concern here. By varying the aperture you do avoid effects from dark current and potentially Raw converters that are not really giving the same type of Raw data at all exposure times. Unfortunately at the same time you introduce a variable in light uniformity over the selected area due to vignetting. You can minimize that effect by taking only a small crop at the center of the image (assuming the light fall-off is symmetrical). But the smaller the crop becomes, the larger a few outliers will influence the statistics. Another potential source of non-uniformity is that there may be a slight asymmetry in the aperture that the (partially sticky) blades of the iris leave open, and there may be a less than perfectly linear progression in the amount of light that's let through (narrower aperture also require longer to close, hopefully to a repeatable final position). Another issue is that at apertures wider than approx. f/3.5, the analog gain of some cameras seems to be increased ...!
I do have one concern with this test, however. I've noticed that SNR holds up remarkably well as resolution decreases, if the noise is big enough.
I'm not sure which resolution you have in mind, so I can't comment on that.
All in all, a very useful exercise and helpful presentation, and it (as usual) learns us a lot about our tools. Do take my remarks a encouragement and not a criticism, because I know how much time it takes to do these things right.
I don't know if Iliah Borg happens to stumble across this exchange, but I would love it if a G1 vs G2 metric could be added to RawDigger. Just a number giving the mean of the G1-G2 differences in a selected area, and the standard deviation of the result divided by Sqrt(2). A large non-zero mean indicates a calibration issue if measured of a uniform patch, and a smaller standard deviation than that of the G1 or G2 sensels suggests the presence of PRNU or of uneven lighting that injects an upward bias in the noise statistics.