It's a computer running a long series of tests directly on the lens while it's is actually focusing and hitting an/or missing target, storing the information in the lens CPU! There is no Perhaps in that sentence.
I haven't seen all the service centres and what equipment they use, hence "perhaps." I also haven't seen all their worldwide service people at work to see what their personal standards and work ethics are. There is no way I could know without going there and seeing (and even if I had done that my observations would be suspect to human observation error).
Nikon calibrate the body completely independent from the lens and vice versa.
Such a procedure would not account for body+lens cross effects, which are the main reason focus fine tuning is needed (as far as I'm concerned). If Nikon calibrates bodies and lenses independently at factory (which they should, of course) then there would be no benefit from repeating that kind of calibration afterwards, unless the equipment is damaged. However, calibrating body and lens independently is not sufficient for accurate focusing (of wide aperture lenses in particular) and calibrating the particular body and the particular lens together is needed. With the D800, I had standard deviation of 9 points on the focus fine tuning scale from lens to lens (after AF calibration of the body at JAS tekniikka). They told me to come back with the lenses if there are still problems, but as I said my lenses are in use. With the D3X the std was only 5 points. This illustrates that lens to lens variations in optimal fine tune depend on the particular body and are not something that can be dealt with independent calibration of bodies and lenses. The D810 however is closer to the D3X in variation from lens to lens requirement for fine tuning (interestingly, it shows less color temperture dependent focus shift but more zoom focal length dependent focus error).
I disagree with Nikon's policy regarding some aspects. For example they don't calibrate f/1.4 lenses wide open, whereas I would only want the wide open image to be considered for calibration. Now, it is well known that focus calibration wide open is problematic and introduces a host of issues but since I would use the lens wide open 95% of the time this is what I require.
Anyway, no amount of calibration will completely get rid of focusing errors, as there is quite a bit of shot to shot random variation in DSLR autofocus. This is easy to demonstrate with a tripod and stationary subject; not all shots will be focused alike even with an easy target. These random errors tend to be more significant at long distances (Nikon even notes in the manual if I recall correctly that autofocus may be inaccurate at long distances, which is an understatement) - I suppose it is a question of speed vs. precision tradeoff - to be able to rack the lens quickly from infinity to nearest distance, the focusing cannot be infinitely precise at all distance ranges and with increasing pixel densities there is more pressure to gear lenses to allow more precise focusing with some cost in speed (as is inevitable). Which is one of the reasons I am not so hot on high pixel density cameras since the AF hasn't kept up with the development of sensors and so there is a lot of waste. But since Nikon only makes their best camera (for my requirements) with a high density sensor I don't have much choice.
Getting back to the long lenses, the random variations in focus especially in low light tend to increase as the aperture is closed from f/4 to f/5.6 and using a TC magnifies the image and any focus error that might exist. So for use with e.g. 2X TC, the lens's focusing system would need to be redesigned to allow finer adjustment to get optimal results. This is very obvious with the 200/2 specifically at long distances and 2X (even though the lens is f/4 with the TC) - there is not sufficient precision to get consistent results with that TC in the long distances (50m+). Perhaps it would be nice if Nikon would make it possible for the lens motor to run at half speed (or quarter speed) when used with a TC to account for the shallower depth of field and consequent requirements.
I would not be surprised if the FL series (and other new lenses introduced in the 36MP+ era) have been designed to autofocus more precisely to account for the increasing demands by people who inspect the image at the pixel level instead of just making practical sized prints (I think we all do some pixel level evaluation to determine which shot has best focus).
Anyway, so long distance performance can depend on:
1 lens MTF itself (and lens+TC combined MTF) - due to atmospheric effects the image quality can fall perceptably but it's better to start with a sharp lens allowing the user to slightly shift the point where the combined lens MTF + atmospheric degradation become perceptible in the final image
2 precision of the AF (i.e. how fine adjustments to focus can the camera control on a particular lens in the long distance range; a TC degrades the precision because of light loss as well as by decreasing DOF)
3 AF calibration of body and lens (i.e. how much systematic bias there is in focus i.e. back or front focus)
4 light level (contributing to 2 precision of the AF in particular)
and probably some other factors that I'm not just thinking of now.
