Hi David,
I guess Patrick might add here some more information, as we all have our very own way to do things as artists.
However, the manual suggest the following:
The suggested workflow for shaping characters:
-> Roughly model the character
-> Sculpt
-> bake (in order to continue work with the low-res object; all details will be included when rendered)
-> rigging
-> texturing
-> animation
You find as well the note that the freeze option is needed to be active, before the Phong Tag is in use again.
My idea would be of course to texture after the sculpting process. The sculpting work does not add polygons to the area that you are working on. All or nothing can be increased or decreased. This results in relatively larger polygons (segments) where you might pull a lot away from the original mesh. This might decrease then the resolution on the surface, and with that stripes become there (for example) larger. Which might be wanted, but I guess in the most cases not at all.
In opposite to this might be the idea, to use a given texture and use the texture as guide to sculpture with it. The texture becomes perhaps distorted, but it might help as well to sculpture at all in the fist place. Then after all is done, BodyPaint 3d might be used to get the texture back in shape.
I used these two examples to illustrate two cases where you start differently, based on the way one likes to work. But on the end, so my current experience, the texture should be done (or fixed).
Again, I believe that everyone should find his/her workflow, that creates the best conditions to work in, but an understanding how the mesh is manipulated needs to be given, to maximize the success.
I would suggest to start with a simple object and tweak it differently: nearly no change and partly extreme change of the prepared source geometry, starting with texture on it, and then texturing again after the work is down. Compare both, in terms of workflow and result. Best practice is mentioned above (manual) but feel free to make you own explorations.
The idea what the source geometry should look like is as well discussed, as the best mesh consists out of square polygons, the targeted shape should be roughly given, to avoid (as mentioned) an extreme uneven size of the final mesh. (Also here, each project has different demands, and if you need nearly none details on the e.g., body, but on the face, prepare the pre model accordingly.
All the best
Sassi