Originally posted by knoxploration the image sensor is centered -- otherwise the available range of compensation adjustment wouldn't be the same every time, as it'd depend on where the starting point was.
Theoretically yes. But the compensation range is about 400 pixels (2mm), the "misalignment" as stored in the profile is about 50 pixels only. So, if the manufacturing sensor centering calibration allows for +/- 50 tolerance, SR and the compensation mechanism would still work good enough.
I am not convinced that manufacturing tolerance is much less, like +/- 10 pixels. Otherwise, I wouldn't be able to understand that quite a few K-7 have been shipped where the horizon wasn't calibrated properly.
And IF tolerances are in the +/- 50 tolerance range, one should probably edit the profile made for another camera and lens.
In order to find that out, we have to look into profiles for different cameras and watch the alignment values. Preferrably for an identical lens, shared between owners of several bodies.
UPDATE:
I looked into some other profiles. E.g., for Ishpuini's camera, the values are all over the place. So, it is lens specific really, not camera-specific. I looked into other people's profiles as well. At 16mm focal length, the value generally seems to be 0.5 exactly. At larger focal lengths larger than 20mm, it starts to deviate. The deviations can reach 10% (0.55 rather than 0.50). So, the interesting test would be two have an identical body/lens combo with 2 bodies and 2 lenses. And measure both possible combinations of body+lens.
But first, let's see if the difference is noticeable and if LR doesn't do this all by itself. After all, a profile contains an author's name. BTW, when editing the file, the file length shouldn't be altered. At the end, there is a white space padding section.