Hi Shana-

It's usually a good idea to do your preprocessing in afni_proc.py or uber_subject. This ensures that the projecting of the data to the surfaces happens in a fairly accurate way (before smoothing) and also allows you to smooth on the cortical surface! There can be some sizable differences when you do the surface smoothing compared to the more "traditional" in volume smoothing. Some comparison and instructions are here. And a general overview of some surface analyses are here.

Importantly, you won't have to re-run Freesurfer! Once you have the Freesurfer folders, and run @SUMA_Make_Spec_FS, you're done with Freesurfer! So really you're just waiting for the functional data to be preprocessed again! And yes, Freesurfer takes a while to run. Though the new Retina iMacs seem to get it under 10 hours. The other best solution is to use something like GNU Parallel (or an actual cluster if you have one) to parallel process the data. I wrote up a way to do this on my blog.

Hope this helps! I can't say how convenient using afni_proc.py is for surface based analyses! Let me know if you have more questions!

-Peter