There are 3 slightly worrying issues about this camera that spring to mind.
(1) No Live View for ultra accurate manual focussing. However, I can see the reason for this. Without image stabilisation in the lens, Live View is not as useful unless the camera is on a tripod. That's one disadvantage of the anti-shake sensor compared to the lens image stabilisation.
(2) The f stop latitude due to the anti-shake sensor is described as 2.4 to 4 stops.
What do they mean by this? What causes such a variation in latitude? It is assumed that, whether or not anti-shake is activated, the final result will depend in part on the amount of camera shake present at the time the shutter is pressed, whether or not that degree of shake is due to a Parkinsons disorder or the motion of a moving vehicle along a bumpy road.
If the IS specification refers to a latitude, then that implies it is in relation to whatever camera-shake conditions apply at the time. If Sony means that the variation of 2.4 to 4 stops is dependent upon the focal length used, then they should say so. Perhaps at either 16mm or 400mm one can't expect better than a 2.4 stop latitude, whereas at 100mm or 150mm one might expect as much as 4 stops latitude.
(3) The noise at high ISO in the images I've seen so far, do not look impressively clean, even aty well below 100% magnification on the monitor.
I'm reminded from previous dpreview comparisons between the Sony A700, A200 and A350 that high ISO noise is not as good as the equivalent Canons. At ISO 1600 and 3200, if noise in the Sonys is as low as the Canon equivalent, it's usually at the expense of detail, and vice versa. If detail is on a par, then the Sony images are noisier.
This worries me a bit. First impression tend to confirm that this trend is continued in the A900. Nikon has already set the benchmark for low noise from a full frame sensor. It is expected (by me, at least) that the soon-to-be-announced successor to the 5D will match the high-ISO low noise performance of the D3 and D700.