My informal testing has shown that f2.8 on my copy definitely does not seem as sharp as f4 and onwards. However, I'm not sure how much of this is actually attributed to the shallow DOF as opposed the lens actually not being sharp.
Should I even expect f2.8 to be as sharp as the smaller apertures or is this unrealistic?
The best way to test is to shoot something flat, like
a sheet of newspaper taped to a wall. The do 4X
blowups of parts in the center and the edges. With my D50 I've found that the weak link as far as sharpness goes is the sensor. There is a low pass filter that blurs the image bonded to the CCD. I think this is why my lenses
appear to have exactly equal sharpness. All of mine of Nikors
Lenses are never perfect even if built exactly
as designed. They will have some un-corrected optical
aberrations. These will always be worse with wider
apertures.
On the other hand there is another law of optics that says a perfect lens will have resolving power proportional to the physical diameter (physical diameter, or focal length over f-stop) This is cause by the wave-like nature of light.
The two rules balance together and the result is a curve. Typically it peaks at about one or two f-stops from wide open. But some very good lenses peak at maximum aperture. Canon publishes MTF curves for their lenses. I wish Nikon did