I've run it just now and got:
-- no warnings for correlation matrix (cut = 0.400) --
Would it be relevant to mention that I'm running this on resting-state fmri data? I've used the -GOFORIT and -Rwherr options and am using the output from that as the final epi dataset. Is that an appropriate usage of 3dREMLfit for removing temporal autocorrelation in resting state data?