You're correct that the main finding reflects flawed statistical assumptions made by the major neuroimaging packages (or so these authors contend), but they did

You're correct that the main finding reflects flawed statistical assumptions made by the major neuroimaging packages (or so these authors contend), but they did uncover a specific bug in one of the three packages (AFNI): "... a 15-year-old bug was found in 3dClustSim while testing the three software packages (the bug was fixed by the AFNI group as of May 2015, during preparation of this manuscript). The bug essentially reduced the size of the image searched for clusters, underestimating the severity of the multiplicity correction and overestimating significance ..."