I've noticed that when this lens is good it is really pretty good, but that there are a disappointing (about 1/2 or more) proportion of longer (70mm and up) shots where the results are possibly useable but not quite as sharp as you know they can be.
My shooting method has varied. Always with VR on (normal not active) but with the auto ISO setup at either regular, fast or faster settings. I began to suspect that the VR on this is not as good as for example the 70-200VRII which is almost always super sharp,
Did a bunch of quick shots in very real world conditions today, mostly just jpeg to make it quicker.
Method: choose a scene (some far, some middle some distant) and with the camera set to manual and the lens to 120mm and f5.6. Let the ISO fluctuate to get exposure right. Auto focus on the sort of scene elements that won't get tricked then switch to MF. Take ten frames at 1/125th with VR on and ten frames with VR off and 1/500th.
1/500th is two stops over focal length whereas the claim for the VRII in this lens is that it can give you 'up to four stops' improvement, which would be equivalent to shooting a frame at 1/125th and getting the stability of 1/2000th.
I don't think so.
The shots at 1/500th with no VR were not all perfect but they were all between usable wonderful. Whereas almost none of the shots at 1/125th and VR on were wonderful and by no means all were useable.
I alternated sets: some VR on first sequence, some no VR first. I shot some with the camera kept to the eye, some with the camera resting by my side between frames. Some were pointed upwards and some straight ahead. etc etc etc. And I think I have proved that the VR is not as useful by any means as the use of a simple high shutter speed.
Next test will be a hybrid: try VR on with a shutter speed of one stop better than 1/f and compare to 2 stops better with no VR.
I have long suspected that some VR systems turn a four or five pixel motion blur into a one pixel blur, but also turn a no-pixel blur into a one pixel blur. That doesn't seem to be the case with the 70-200, which also has VRII, but I do expect to find that the best way of getting optimal results on the E with the 24-120 will be either tripod (in which case I'd likely be using a good prime) or to use two to three stops of shutter speed better than 1/f.
I will report back! And would love to hear if anyone else has an opinion or experience here.