Using masks to select areas for noise reduction seems like a good idea, thanks! In the straight raw conversion the field in the foreground is full of colour noise but somehow the trees in the background do not have as much of visual noise that it would bother me. In the DXO NR processed image, the forest looks too processed but the foreground looks okay (to my eye). So I could use a mask that combines the field from the NR processed image with the forest background without noise reduction, or some intermediate mixture. However, for me it is important for the image also to look decent in a print and I suspect the noise would become more apparent there, so some processing of the forest as well is probably a good idea. I can always experiment.
However, I don't see the problem as so much related to any specific camera - ultimately this type of noise mostly comes from the fact that only a small number of photons arrives on the sensor during the exposure and since individual photons arrive at random times, there is a lot of uncertainty in how many get recorded in the given fixed time interval, leading to a lot of visually apparent noise. What noise reduction can do is try to make the image look more pleasing to the eye. In very low light, the human eye utilizes rods which do not have colour discrimination ability, so in a way it is okay for the image also to become more monochromatic. DXO PRIME seems to process out the colour noise and leave luminosity information within the deer fur for example. It kind of makes sense to do that as it maintains the overall colour of each object and detail within each object is more monochromatic. I guess this creates a little bit of a painted appearance. Still it looks a lot better than the unprocessed original in this case.
