I am not an downsizing maniac.
Nor am I. I generally upsample more frequently than I downsample when making prints, but recently I've been doing a lot of downsampling in order to display jpeg images on my plasma TV. I would say that it is very rare that I would display any image at its native resolution, whether on monitor or print, although in practice I frequently hand the resampling job to Qimage.
I see foolishnes in downsizing presharpened JPEG images (probably after noise reduction) for comparison.
Well, that's where you and I differ. It might be regrettable that the only images available for comparison are in jpeg format with an unknown processing and sharpening already applied. However, to refuse to equalise the size of both images on the grounds that they are jpegs, makes no sense at all to me.
You should either discard the images on the grounds they are not suitable at all for comparison (in which case it raises the question as to what IR is doing in providing images on their Comparator which are not suitable for comparison) or you do the best you can, despite the fact that the source images may not be ideal.
Does it? I don't know about that.
Yes. The downsampled D3X image is clearly sharper and more detailed than the D3 image, whether standard bicubic is used or bicubic sharper. The situation might change with different processing of the RAW images, but I suspect the bottom line will be the same, ie. the D3 has no IQ advantage over the D3X in respect of resolution or noise when comparing equal size images.