The sharpening in PM 5 is the same as previous 4.x versions. It is essentially a 0.3 radius unsharp mask, 200%, and threshold of 2, but the sharpening is only applied to the luminance and the chrominance (which is typically more noisy) is left untouched. (For a grayscale image, which obviously doesn't come out of a camera, there is of course no chrominance.)
For testing, I took a D800 JPEG photo at 400 ISO and did a Save As with sharpening and no scaling at quality 90. I did this for both PM 4.6.9 and PM 5. Then I loaded the two saved JPEG files into Photoshop and subtracted one from the other, and the standard deviation was 0.9, meaning that 2/3 of the pixels are within one level (out of 256). A handful of pixels differed by as much as 10 levels. But comparing the images directly (by pasting one on top of the other as a new layer then hiding/showing the layer) at 200% magnification I could not tell visually where there was any difference.
So although the sharpening has NOT changed in the Save As operation (I know because it is my code), we are using a new JPEG library implementation. This would explain the slight difference between the two versions of PM due to round-off errors in the math.
Now if you are sharpening previews and you are looking at a noisy photo that has already been sharpened, then yeah, I expect it to look over-sharpened. There is also the possibility that the caching system is doubly sharpening, but this would not affect a Save As operation (unless of course you are looking at the sharpened Save As file in PM and sharpening that).
If you can devise a scientific way for us to verify that there is a big difference then please let us know how to reproduce. As far as I am concerned, this is all a subjective discussion without a way for me to verify any difference between the two versions of PM.
Photo Mechanic is internally using the code from the original Camera Bits plug-in "Quantum Mechanic" and it is capable of controlling the sharpening radius, amount, and threshold, as well as noise reduction (chromatic or luminance de-speckle), but we have not "exposed" the parameters to the user interface yet. Its on the long list of features to add.
Regards,
--dennis