OK... so...
Firstly, thanks for all the interesting input and discussion. You good folks never disappoint when I have a query or I'm looking for opinions
I spent some time this morning running a few tests. For the same 35mm Ilford XP2 Super 400 negative, I digitised it first with the K-5 + DFA100/2.8 at f/5.6, f/8 and f/10, then with the K-3II in non-pixel-shift mode, and finally using pixel-shift. I loaded the files into RawTherapee using AMaZE or pixel-shift demosaicing as appropriate, without noise reduction or sharpening of any kind applied, then opened them in GIMP 2.10 as 16-bit TIFF files using RT's external editor feature. I cropped the borders of each image, inverted the tone curve, then automatically adjusted the input levels to attain almost-identical black-point, white-point and gamma for all six files. I then examined each one at 1:1 reproduction. As a final comparison, I scaled the K-3II files to the same pixel dimensions as those from the K-5.
Looking at the K-5 files, there's little noticeable difference between the f/5.6 and f/8 shots... but - and I say this while chowing down on a large slice of fresh humble pie - you folks were quite right... the f/10 files are slightly softer; enough that I
will use f/8 by default in future, unless I consider the additional depth-of-field to be essential on a case-by-case basis.
Comparing the K-3II non-pixel-shift files, they benefit from greater acutance than the K-5 files - and without any unpleasant rendering effects in the grain, with this negative at least. The scaled-resolution files also look better, to my eye, than the K-5 files - though both are very good, and when the K-5 images are sharpened just very slightly, it's difficult to tell the difference. Still, the K-3II definitely produces sharper results, IMHO.
Which brings me to the pixel-shift files... and, rather surprisingly, I don't see any obvious improvement over the non-pixel-shift versions. It's possible that the film is the limiting factor here, though, and I'd like to try again with negatives from a low-grain, high-detail film. I've not shot any such film yet, so this is something for the future.
I'm still very happy with the results from the K-5, especially at f/8 - but since I note a definite improvement and nothing undesirable from the K-3II, I may switch to that (or my K-3) going forward. I can see how it could make a difference for larger print sizes.
Thanks again, all