Good to know. Never had to do this before, but I found some online websites which assist with the jpg to xmp conversion, and I was able to import that into darktable, and apply to a copy of the raw file.
Would you know of any desktop app/tool (maybe darktable itself) which does the jpg to xmp conversion?
Yes, there’s clipping in the clouds, but since there’s no colour or fine details, they can be recovered relatively well.
You just load the jpg as if it was an xmp…no conversion …DT will pull the history stack …you may have to change the file selection box to all to allow you to select a jpg instead of an xmp’
Loading the JPG as a sidecar, I seem to be posting this quite often: darktable 4.0 user manual - history stack, under ‘load sidecar file’ (just switch the selector from *.xmp to All files):
Images that were exported by darktable typically contain the full history stack if the file format supports embedded metadata (see the export selected module for details of this feature and its limitations). You can load an exported image as a sidecar file in the same way as you can with an XMP file.
I think the loss of detail in the grass on the left is apparent even on the screen, without pixel-peeping.
I do love a challenge, if it re-enforces learning, as in this case, or a nice break from the usual.
Regrettably, there’s only soo much one can do to challenged images, with digital tools, Each fix, creates new problems.(side effects), somewhat like pharmaceutical medicines.
Occasionally, not a problem. The more I work with images, though, the more I cull, before bothering to invest any time to bring out the best in the few images that are worth the effort to edit to perfection.
Could this area on the left, already be slightly out of focus, optically. cos you’ll find in all the various versions, on the thread, including the OOC jpg, that left side of the grass is consistently less distinct than the grass on the right.
With local contrast turned off, there are no modules settings which could have introduced (added) any defocussing or haziness, so the observation on detail, may be one based on the actual image/rawfile itself, and there’s a limit to what can be enhanced/repaired in this aspect.
Of course an attempt can be made to artificially sharpen it, but that has its own side effects - no free lunch, only trade-offs.
Awesome. Thank you.
I am reading that manual, or rather, the one for 3.6 which I’ve switched to, as fast as I can. So much to cover…
Sure, it’s not a great shot – it’s a family snapshot, I kept it because it’s us in the photo (it was the least bad of the few shots taken at the same location).
I also tend to cull more as I age. 
But still not enough. 
That is what I like about it. To me, good enough + family = great.
I took another stab at this. I started from a preset I have been working on for RT 5.8 Dev, and after I applied it, I left the background untouched aside from some exposure comp. I then worked on the family with some local edits. I tried to keep the lighting as natural looking as possible while still bringing out the details.
On Darktable 3.6
Tried to get along without masks, which however might be the best solution for this photo, I think.
Nice work keeping the background details and recovering the foreground. Somewhat overly contrasty for my taste, but tastes vary — and it’s a rather difficult image.
Thank you for your comment. I agree, that is too contrasty.
Here is my second attempt. Maybe better.
Very nice! I’ll definitely check out the xmp later.
Thanks.
For the export, I used the lua script RL Output Sharpening, that you will not find in the xmp 
My fun with GIMP Layer copy Screening mode, again layer copy Screening mode and change opacity, now LAB Colors
