The posts above explain it pretty well. To summarise from my point of view:

A single morph won't work as you know, because of course the path of the zipper has to follow the curve of the leg (in my case) or the back (in yours).

I decided to simplify the opening to five discrete stages: fully closed, and quarter, half, three-quarters and fully open. I modelled each stage in the cloth room by constraining more vertices each time. The actual moving part of the zipper wasn't involved at this stage - I placed it in the appropriate position by hand afterwards. It took some jiggery to keep the different stages morph compatible with each other.

However the big decision was to use geometry switching rather than trying to link a lot of limited-travel morphs to one ERC dial, on the basis that it made things a whole lot simpler. Because each of the unzipping stages has the same number of vertices, in the same order (this is not a requirement for geometry switching), styling and fit morphs that I made will apply equally to all stages of unzipping.

Notice also that each boot consists of only one group: lFoot or rFoot. A dress could be more challenging, depending on whether the zip needs to travel across a group boundary or not. Linking geometry dials using ERC is technically possible, I believe, but the idea is making my mind boggle somewhat.

Edit:

Just to address a point from earlier in the thread: I didn't bother modelling teeth for the zipper. It's just a flat plane with an appropriate texture map applied. For anything other than extreme close-up, it looks fine.