hi Michael
I can't answer all your questions but I do have a few things that might help in the meantime. If the preprocessing done in SPM included smoothing then you need to account for that in your 3dClustSim command.
The data come off the scanner with a bit of smoothness, and the smoothing kernel used adds to this. This smoothness increases the likelihood of finding sets of contiguous voxels that are above a threshold and so this is why 3dClustSim needs to know this value. 3dClustSim isn't blurring your data, it's blurring a noise volume to have the same smoothness as your data.
As I understand it, resting state data basically are the errts of a regression without task predictors (at least in AFNI) so they should be approriate for 3dFWHMx. I'm not sure about how much of a difference detrending would make in this case, perhaps someone else can chip in.
Finally, you can average the a b c parameters from 3dFWHMx -acf across your subjects and feed this average to 3dClustSim.
Hope this is helpful.
James