The only valid way to evaluate noise reduction and sharpening is to view the image at 100%.
Anyh other view will be misleading.
Different programs use different algorithms to scale the image to fit it in a window, so noise reduction is almost certain to look different.
Thanks! Here's a version where LR is set to 1:1 and Paint is set to 100%.
My expectation is that the rendered image will be nearly identical to the on screen edit. It appears the white balancing and clone/heal all rendered well. Whether the grain is the result of misapplied noise reduction or noise / grain added during rendering due to some other reason is difficult (for me) to determine, but is that basic WYSISYG expectation unrealistic?
Now - on to some very weird results (happening as I write this)!
Here is a 1:1 comparison from LR Developer and the rendered version with Paint at 100%
So, I thought, I would try moving it to PS CC with all edits in and compare / render from there. That comparison is below. That is in line with my expectations.
However, I used MS Photo Viewer on this comparison, so I thought I'd do a fresh compare with LR using MS Photo Viewer. When I did, I somehow jacked up the edits on the sunset in LR, so I readjusted and re-rendered it. You can see from the histogram the color settings are a bit different. NOW the LR noise render is much closer to expectations. Why?
Here is a side by side of yesterday and today from lightroom. The export settings did not change. I DID do an export at 100% luminance accidently and I opened PS and messed around there a little. Still, when I backed LR back down to 78% (yesterday's setting) it still rendered well. The color setting in HSL are still a little different, but I can't understand why the noise seems to have cleared.
I will attempt to recreate the issue.
Don’t use Paint.
Reimport the exported image into LR and use Library X|Y to compare the two, zoomed in to 100%.
I started to reply to this and got pulled away for work. The process of rendering and importing the render to do a "candidate" comparison at 1:1 really helped nail down the issue. That allowed me to test the options in a more predictable way.
I'm really not sure why my first renders were so off base now. Maybe I had something messed up in the export options initially and just kept making things worse as I struggled through the options. But in the end, these last two comparisons (comment 4), the grain was caused by sharpening.