I'm running fslcc to assess correlations between 670 volumes (probably a bit more than usual?). I run it with the noabs option as well as thresholding at 0 (so that all correlations would be recorded). However the results don't end up with the expected 448900 rows (670 x 670) - after running it in an identical fashion three times using the same command I've ended up with 448898 rows, 414200 rows and 423923 rows. Different columns get skipped on different occasions. I have a hefty amount of RAM so I don't think that's an issue. Has this ever happened to anyone else?