Paul, you summed the lines over lines of my question up into one sentence: “is the quality of fitting in the first 400 time points different than that of the remaining ones, such that the residuals would have very different properties?”.

Thanks for this one. :D

This is basically what I was thinking about. Normally, I don't use bandpassing and leave the polort option to the automatic standard calculation. I only used

    -regress_polort 2         \
    -regress_bandpass 0.01 1

on this particular dataset because my laptop never seemed to finish preprocessing the first subject in the pipeline when AFNI set the polort option to 34 by default. Rick then suggestet to use the bandpass and polort settings above.

Just as a side note here: with the default polort option and no bandpassing, the file folder size of the AFNI proc output for one subject was around 30 GB (at the point where it was detrending the data - baseline funcs ...), and now, using bandpassing 0.01-1 with polort 2, the final folder size per subject reduced to around 12 GB.

As for your answer: thanks, this helps me out. I will also make a comparison for one or two subjects, just to be sure that the results of cutting before vs. after AFNI proc look the same.

Philipp