Hi Zoe, 

I'm not an author of this fantastic toolbox, but maybe I can help you. To answer your first question: this really depends on which steps you have run/would have run in Conn. You can load your individual structural and functional files (as preprocessed as they are) and, when you get to the preprocessing step, select any preprocessing steps you would still like to run/or de-select those you'd like to skip. During preprocessing and denoising, the program will automatically add first and second level covariates (such as realignment and scrubbing parameters, or quality assurance), which you'll need to add manually if you skip those steps in conn. 
To answer your second question: if I remember correctly, in my analyses, I usually didn't have to run the preprocessing again, but the first and second level analyses. You can speed this up a bit by selecting "skip already processed subjects/seeds ..." in the window that will appear once you have changed your ROIs. 

I hope this helps, 
Best wishes, 

Charlotte