[Shiraz (Ray)]

dark current/dark noise is difficult to measure. As you say, there are two populations - well behaved pixels that generate the basic dark current, plus the lower number of abnormal pixels, including locked-on ones and warm/hot pixels that generate more dark current than normal ones.

the abnormal pixels can almost all be removed with stack rejection when you are putting together an image, so it is probably appropriate to ignore them when measuring dark current. I find that the mean of manually selected small regions (without too many warm/hot pixels) works OK, or you could use a non-linear measure that accounts for outliers, such as Median.

Dark current within a pixel can be averaged over many dark subs and that will give you the underlying dark current in each pixel. However, when you take lights, you also get dark current and the value you get in a pixel in one light will be different from that in the next light due to random fluctuations in the dark current. You can subtract the average dark value for that pixel to get rid of it in any sub, but you will still be left with fluctuating values in successive light subs due to the random fluctuation. This is dark noise and it will generally have shot noise characteristics. It can be reduced (relative to the signal) by longer total integration and you can increase the total integration time either by stacking a few long subs or lots of short subs. Provided the short subs are not so short that read noise becomes a problem, you will get identical signal to noise from either approach. The cutoff point for how short the subs can be is fairly flexible and determined by numerous factors (including pixel size, fno, read noise, sky brightness, dark current), but for the low read noise cameras you are considering, you should be able to use subs in the 2-5 minute range for typical optics - at least that would be a good place to start experimenting.