Upload & Sell: Off
I guarantee you that there is more than 36MP of information in a technically perfect K25 slide. I used to play with drum scanners back in the day and I found that the optimal scanning resolution was roughly 6000dpi. Any less than that and you were leaving detail on the slide. Any more than that and you were just making bigger files with no additional info. At least that was my experience.
I've been tinkering with DSLR duplication since 2010, using the 5D2 until quite recently. In some cases it produces inferior results to slide scanners such as the 5000ED (which I really need to sell now that they're worth a fortune and I never use mine), but in other cases the results are nearly indistinguishable. The thing is, when you have 50,000 slides to digitize, scanners just don't cut it. I can dupe a slide in about 10 seconds using a DSLR. Like Lloyd (although for vastly different reasons), I wish I had a mythical 60MP D4X to use for slide duping.
But as I said, the D800 experiment was largely a failure in terms of making any improvements over the 5D2. However, your point about dynamic range and shadow performance in real-world images is very well taken. Having been a Canon man for many years now, I'm dumping my entire collection based solely on the phenomenal strength of the D800E images and Nikon wide-angle glass. So even though my original reason for switching was a flop, I'm ultimately making better images than I ever could with my old gear.
Just FYI, there is a lot of info on DSLR slide duping scattered all over the web, although virtually none dealing specifically with the D800E. Also FYI, I tried several lenses on both the 5D2 and D800E. My conclusion is that the 5D2 makes best use of the 100/2.8 macro, whereas the D800E performs best with the 60/2.8G micro. Then again, I don't have a 105/2.8 VR to try.
A brief (and admittedly lacking) view of what I am talking about when I say 5D2 vs. D800 for slide duping detail.