Graynomad:
I guess it's hard to orphan 500+ existing shields, but if some/many/most of them won't work anyway it doesn't matter.
Exactly. No-one is going to "throw out" their old shields just because they've bought a Due, anymore than they'll throw out their Unos and Megas. Why would they? The AVR product line still has a future, and the the ARM products complement rather than replace these products. Even if Arduino never made another AVR-based board, the old 8-bit AVR Arduino "ecosystem" would now continue indefinitely under its own momentum.
The problem with the IOREF pin approach is that adds complexity (and therefore cost) to any shield design that wants to use this technique. Cheaper and simpler to design an AVR 5v version and a separate ARM 3v3 version, if needed (I suspect many if not most shields will be suited for either AVR or ARM rather than for both). The fact that a lot of people are going to be buying shields designed for AVRs and finding they don't work for the "Due" is going to cause endless aggravation that could have been avoided, I suspect. I hope I'm wrong.
The reason I believe this is an opportunity lost is that, while anyone can design a dervative board with an alternate header layout, only Arduino is in a position to set a new "standard" layout by simply releasing a new product. So no, remedying the "lack of vision" is not quite that simple.