Originally posted by falconeye @Christine: Your spherical aberration theory is good. But I fear that we have some evidence against it like the dependency on light color or that stopping down doesn't help (beyond a proportional reduction of the confusion circle of course). Maybe, an extension of your theory to include CA or Bokeh-CA would be required. It still wouldn't explain the strong dependency on luminosity though.
Thanks for humouring me, but I suspect I have no idea what I am talking about!
I started building a theory involving spherical aberration in the main lens interacting with chromatic aberration in the micro lenses focusing into the AF sensors, but quickly realised it was just wild speculation as I don't have a good working knowledge of the AF mechanism.
I do believe though that the key factors appear to be light temperature, brightness and possibly lens type (this could cover a multitude of factors including available light, aperture, lens curvature, etc.)
Would be interesting to speculate what the variance is among units. My suspicion is that the problem is more or less present in all units, but in different degrees - hence the inconsistent reports (ranging from no observable issues to severe focusing problems affecting all shots in low light).