Bart, this thought occurred to me the other day: Is there any advantage to doing capture sharpening in the raw conversion as opposed to doing it on an already "baked" demosaiced file? I believe Eric Chan has said that it is preferable to do lens corrections in the demosaic process as opposed to after the fact (though I may have misunderstood what he was saying).
That depends on at which stage the corrections take place. It is e.g. possible to 'do' Chromatic Aberration correction before
actual demosaicing, which would not only improve the color accuracy of the demosaicing itself, but can also help (luminosity) resolution coming out of the demosaicing stage, even before Capture sharpening.
When a relatively mediocre resampling algorithm is used for small resampling adjustments, e.g. lens distortions (especially near the optical center) or rotation/keystone correction, then proper (!) Capture sharpening before the resampling may help to preserve micro detail a bit better. Here
(Rotation examples) is a nice demonstration of the devastation at the micro-detail level that can result from mediocre resampling methods. It's visually even better to make a relatively soft conversion and use deconvolution sharpening to recover some of the detail because it looks more homogenous instead of fading in and out of sharp detail (sharpening that can make it look even worse). Of course it also depends on the actual image detail, because e.g. out-of-focus (OOF/DOF) zones have no micro-detail other than noise.
The key concept is proper Capture sharpening
. Without that, all bets are off. That's why I'm stressing the need for proper Capture sharpening so much, it's at the source of what follows. And garbage-in-garbage-out (GIGO) still rules ...
Without such automatic resampling, it's always better to use a superior deconvolution sharpening method after e.g. ACR than as part of the postprocessing in ACR. It complicates the workflow by adding an additional step, but such is the fate of those striving for perfection.