Turn off highlight reconstruction, and you’ll see. The screenshot above is with filmic disabled and exposure dropped.
This raw white point is probably wrong, that’s why the raw clipping indicator does not show clipped pixels:
If you lower it a bit:
The indicated clipped pixels match up with the magenta:
Hmm… I used the raw overexposure clipping option in darktable and no clipping was indicated, hence my comment. The settings for my indicator in DT V4.7 are shown here. However, then as suggested I turned off highlights reconstruction module and I got a very weak Magenta cast in the cloud, which was barely noticeable (scene shot below). I guess I can at least be happy that DT handles clipped highlights well if it has enough to work with. I love the new inpaint opposed option. The developers did great to include that as the new default method.
These are highly dependent on the raw white value. If you look in exif data the values are 0 for black and 4095 for white. Then an offset of 15 is added to black. In RT it is suggested that 3799 or something plus minus 50. is the Peak value for white and is used for lower iso. and 4095 for higher…in any case dropping the threshold of the raw warning only slightly pops on the warnings … so often there is nuance in these files and it’s not a bad idea to check the exif data just to confirm things
.m