Here is the Canon 7D sensor measurement at PhotonsToPhotos vs. the S1. Not 7D mark II, but the original. Worse dynamic range than a 6D2 or RP.
View attachment 183832
That's a difference of 3.5 stops. Based on that you would assume a massive difference in shadow recovery, would you not?
I've posted this before, but so that it can be easily compared to the Panasonic sample you posted:
View attachment 183833
The shadow recovery in the Panasonic sample is deeper, but not by visual leaps and bounds as people who study graphs and scores might imagine. (And I held the sky
)
Point is once you get to the roughly "1ev" range between the 5D4 and a S1 or D850, you're looking at small differences in noise quality while pixel peeping. You're not actually looking at differences in what can be recognizably recovered and used.
The problem with both PhotonsToPhotos and DxO is that nobody includes a clear legend which puts these differences into a meaningful context.