>Why is it so surprising that they've improved? We don't >really know what they did, and certainly nobody's used one >yet. I would guess that the speed and/or precision of the VR >mechanism has improved by some material amount, and that it's >probably better than what we had in VR-II. I'd guess that in >reality (or at least in my personal use) it probably won't >make five stops improvement, but I thought that VR was a >little optimistic at 3 stops and VR-II was optimistic at 4 >stops.
Exactly. I don't think anyone doubts it improved. I think it's like so many other new versions, each exaggeration has to build on the last.
In a brief look I didn't see any third parties try to scientifically measure it. I wish I had time and equipment to do so, it would be an interesting experiment. Though one could always claim "that's not realistic" for whatever simulation you do, since everyone's motion is different, from palsied older shooters to shooters in a riding jeep.
But really -- 5 stops is x32. That's a big number.
To those shooting at very slow speeds -- I suspect you also have steady hands. The real test is turn it off, see how slow you can shoot reasonably, then go 32x (Or for now 16x) as slow and see if you can still get the same Sharpness. Don't just got 16x below 1/focal length and say "see", because I bet as an experienced shooter you can already shoot 2-4x that slow with good technique.