Rabanito's original full size image is an image that is very sharp, but with shallow Depth of Field, and with some noise.
As was obvious to me from the start, the original reduced size image did suffer from the process of downsampling, with ringing and aliasing/stairstepping artifacts as the tell-tale signs. So some improvement in that initial step was possible. Using a better downsampling method, produced a more robust image for further processing/sharpening.
The amount of subsequent sharpening remains a personal preference thing. But the sharpening method should IMHO preferably be either deconvolution sharpening (Focusmagic is still a great tool for that), to restore some of the resampling blur or, as it has started becoming available, detail replacement by AI.
Deconvolution, or AI, both allow minimizing the risk of amplifying the artifacts that were already there, or of introducing new artifacts.
It was an interesting exercise.
Cheers,
Bart