Stuck locks after dbrm_backup

It was reported that, on production clusters, table locks remain held after mcs_backup_manager.sh dbrm_backup completes. The symptoms they're seeing: viewtablelock shows table locks present after the backup finishes that were not present before it started.
Subsequent cpimport and caldroppartitions operations on the affected tables block or fail.
The locks do not clear with normal write activity — they require manual cleartablelock <id> intervention.
In some cases the locks reappear after a BRM/controller restart even though they were manually cleared earlier.