Hi, Robin-
I will comment on a subset of points, and leave it to finer minds to reply about the rest.
Re. the SSW part:
+ in short, yes, the @SSwarper results are typically worth it. Note that I have been processing groups of subjects here on our Biowulf machine, using 16 CPU threads, and most brains took about 70-75 mins, though a small percentage of stragglers took much longer (~2.5 hours). After a while, adding so many CPUs might not result in much benefit; I guess I could/should test this here to see what a good number would be for balancing speed/usage. But I typically use 16 CPUs (or thereabouts).
Re. Q5, 3dAutomask+deobliquing:
Looking at the proc* script from a Bootcamp example, i think that align_epi_anat.py by default will automask the EPI before trying to align it; therefore, before any deobliquing specifications have reached it, an automasking is performed (and using @SSwarper -deoblique only affects the anatomical vol, not the EPI). I don't believe that this is a "scary" warning in this case (but I also don't believe it is unexpected, even with those specifications).
Re. Q0, check-flipping:
I am highly curious about that error. I will run the check flip on the dset you uploaded and see if that provides the same error-- I have not seen that previously.
Note also that the APQC HTML output provides images for you to quickly check and visually verify the flipping results-- the only real way to tell. Sadly, human eyes have not outlived their usefulness in data processing, despite what you read in some corners!
--pt