I think we are basically saying the same thing. My mother tongue is Italian and I used the term “should” with the meaning “recommended for optimal results, but in no way mandatory”. I totally agree that offset isn’t critical at all and that is “has to be set high enough so that the histogram is shifted to the right far enough to avoid clipping data”, which is basically my recommendation (while others stick with formulas like offset= gain / 2). Full agreement on the QHY163M quality too.
I have a different experience regarding offset value at medium-high gain: I found that a narrow-band long exposure (10’ or so) at gain 200 requires a slightly higher offset (in the 50-60 range), but that could be attributed to variance among single instances of the camera/sensor. Regarding the dark library, as long as you stick with a single offset value for each gain value you’re going to use, you’ll have no issues. By the way, I think most users select just two or three gain levels, so you need only two or three dark frame sets for each exposure time and each temperature.
@buzz Unfortunately, this camera shows amp glow in long exposures and it also implements black point compensation. This means that you cannot scale dark frames, as they are not linearly related to exposure time, so PixInsight’s optimization is doomed to fail. By the way, for this camera I always recommend to calibrate manually (i.e. avoid using Batch Pre-Processing script) and to disable dark optimization. I wrote a paper/tutorial on this topic, so if you’re interested please let me know.