The argument with the axis: why is the object in focus not affected, but the not in focus are?
Wouldnt this change the axis on the in focus object too?
Well, thats the magic of the IS
The angle of the plane of focus in relation to the object/camera may be affected, that depends on the way the IS works. If the IS works as a shift element, then the focal plane is not affected. If it works as a tilt element, then there is a minimal change:
If the maximal correctable shift is some 1mm in any direction (my guess, maybe its lower), the subject distance is still 1m, then the angular change of the focal plane is:
0.001/1=tan(alpha), solve for 2alpha, which is approx. 0.002rad or 0.115 degree.
I have no clue how an IS really works...
But more important is the actual position shift of the objects out of focus:
The position change of the far object can also be calculated: 2mm/1000m=x/10mm => x=0,2mm (32 pixel shift on a 5DIII, or some 46 pixel on the 18MP crop sensor as found in the 600D and others)
These are purely theoretical values, as the IS element is surely near the nodal point to lessen this parallax error.
I try to think of CR the next time I get a bad pic due to IS interference.
If you want to force bad bokeh, try the 70-200 or the 24-105 at its maximum focal length, focus on something like a single reed, a flower/fruit in a tree - something that is isolated, small so you get near MFD and has anyway a wild background. At least all pics i got with this IS-bad-bokeh were on such a setup - normally they were visible on the camera LCD so I took it again.