I agree, there is a co-mingling of issues here. One should analyze one thing at a time for correct scientific procedure. I am doing some very elementary stuff with the images which Bart so kindly shared with us and I'll report back if I have anything useful to contribute as a result of that.
I'm back (ugh! some may say - too bad )
First, I like Bart's basic image, because it has the kind of micro-detail and frequency which lends itself to the testing one needs for drilling down on the issues discussed here.
Second, there are five things going on at the same time relative to the raw file: (a) The file format has been converted from CR2 to jpeg, which destroys a huge amount of data right from the get-go; ( the bit depth has been shrunk from 16 to 8 which destroys a lot more data - so we're starting the examination from a compromised file. © There are two kinds of downsampling, and (d) there is sharpening versus no sharpening. (e) In this particular file, there is a strange underlying phenominon with the brickwork on the windmill (only) which shows under some conditions and not others: this is a pattern of concentric bands accross the brickwork. These got created and embedded somewhere along the process between capture and processing, and, as I say, they show under some conditions but not others. I would expect they are completely absent from the original raw capture, but got introduced with one of the above-mentioned manipulations.
Starting with Bart's 25MP file, the first thing I did was to examine what shows on the display (LaCie 321 resolution 1600*1200) as a result of resampling. The first resampling I did was only to change the resolution from 300PPI to 72 PPI (quite a dramatic change, but what one would do making high-res images useable on the internet) without touching the linear dimensions. This downsampling, viewed at 100%, produced no obvious impairment of the image. Then I added to the change of resolution, a downsizing of linear dimensions from the 12*18 (roughly) Bart sent, to the (roughly 7.4*11.1 which he used in his other renditions. The combination of the change in linear dimensions with the change in resolution brought out factor (e) above.
Next, I printed (in my Epson 3800) all three of Bart's low-res JPEGs on a large sheet of Epson Exhibition Fiber paper at the same resolution as provided (72 PPI) and did absolutely no adjustments period, so as not to introduce yet more variables into the stew. Of course, the resulting print is somewhat pixellated, as expected at 72PPI. Therefore to examine for other impacts, one needs to kind of look through the pixellation, but in a way perhaps this is not so bad a thing, because it does allow one to TRULY PIXEL PEEP - and this IS what we're into here: sheer pixel-peeping. So, first I examined the two unsharpened low-res JPEGS (i.e. the one downsampled by what Bart calls "the proper way" and the one using Bicubic Sharper. Even with a 5x loupe, the printed images show no difference of quality, except for factor (e) above - the bicubic sharper image showed it quite mildly and the "proper" one didn't. That doesn't necessarily mean the "proper" one is "better" - it may in fact be less accurate, but that's unclear.
Then I compared the two images downsampled with Bi-Cubic Sharper - one capture sharpened and the other not capture sharpened. Here again, een with extreme pixel-peeping, the effect of Capture Sharpening is virtually indetectable on paper, and that is how it should be. In a multi-staged sharpening workflow, Capture sharpening is built-upon to get the final effect, it is not an end in itself.
Now, to escape from whatever the confines of pixellation, I reverted to Bart's 25MP 300PPI JPEG which has a - to us - starting file size of about 12*18, and I made a copy of it. I left the original intact, and for the copy I simply added PK Capture Sharpen Hi-res Digital Superfine (because this is a very high frequency image). I printed the two images on Epson Exhibition Fiber paper (BTW, in all sets of prints, I used Epson's highest quality settings of Super Photo and low speed, with Photoshop Manages Color and Printer Color Management OFF - took a while to do). This time, we are comparing for sharpening impact only - and here again, even examined under the loupe - there is nothing negative, no halos, and in fact these images are pretty-well indistinguishable on paper, and that is very much what I expected also.
I've now exhausted pretty much what I could do with the materials at hand, and I've come to a provisional landing that much of this discussion is fun intellectual self-gratification, but it has very little, if any, real-world significance.