To stitch images with the HTS, two or more images are taken using the shift function and then assembled in photoshop. The tilt function should be precisely set to zero, because it works in the same direction as the shift function. If there is a slight tilt, it will affect the far sides of the panorama considerably. One cannot rotate the tilt and shift angles independently.
I'm not exactly clear on what you mean with "it will affect the far sides of the panorama considerably".
If you are referring to Photoshop's difficulty with stitching shifted images correctly unless they only require simple shifting to align them, I'd say; use a better stitching program that allows to dial in an offset parameter (per image if needed).
The HTS has an electronic indicator for tilt and shift. Can this indicator be trusted when set to zero or is there another way to proceed when taking panoramas?
I'm not exactly sure how they've implemented it, but it cannot be as accurate as calibrating it on a real scene shot at different settings. Not only will there be mechanical play involved, there is also limited accuracy of the sensors (even a digital calipers also needs calibration before use). Using a longer measurement base, a distant object, will be much more accurate. How repeatable the found settings are for other shooting scenarios, remains to be seen, and I suspect that the post-processing by Phocus will try and compensate for some of that as far as lens distortions are concerned.
Good Pano-stitching software doesn't care about the input, it will handle it well regardless, and it does so on the actual images. It does help that Hasselblad gives the entrance pupil distances measured from the image plane, as a starting point for further calibration (good stitching software will allow to additionally tweak the 'focal length' setting for even better stitches), because that remains an important input parameter for the highest quality stitches.