I was contemplating the 50L fiasco today and had an idea. I wondered if enabling the expanded AF area on my 5D would have any affect on the focusing anomaly. I casually set up the focusing chart and enabled the expanded AF points (custom function 17) while in one shot focusing mode and center focusing point selected. As you might expect no difference.
Then I put the camera in AI servo mode (still with center focusing point selected) and shot a frame without enabling cf 17 then several with cf 17 engaged. To my surprise the shots with cf 17 enabled were consistantly better on the focusing chart.
Now Iam not saying this solves the problem. It only improves the shift slightly on my camera on the focusing chart. I have not had a chance to try it in real world shooting. I was hoping some of you would try to duplicate the findings and maybe try this in some real world shots if you have not already.
This post contains a typical cropped shot from my 5D/50L at f/2.8 AI servo without cf-17 engaged.

Setting cf-17 to on only seems to change performance on the focus chart when the camera is held 

