rephrase of previous comment: The region_mover script initially caches a list of regions to move. If the table is deleted after the cached list is gathered but before all regions are moved, the script can get stuck attempting to move a deleted region.
In a related but likely separate issue – if a new presplit table is created, as the region_mover is emptying an RS, the emptying RS is a candidate for new regions and will get some of the new regions. Ideally this is fenced off so that it does not get regions assigned to it, but this would require some ZK. This seems less important because the majority of regions will be moved off the RS and the few new regions can rely on the automatically fail over to other RS's.