How about merging spatial alignment logic into your brilliant
MatchFrames script? I guess It might boost its efficiency quite dramatically especially if sources of the two clips are very different like VHS capture versus DVD. So it would be easier to reconstruct your old but lovely VHS cut (English?) of a title by replacing its crumbling frames with the ones from a remastered DVD (German?) which is cut, cropped and/or panned differently. This might be the holy grail of poor man's remastering
I'm sorry if I've said something silly but I'm just a hobbyist