Backup Piece Handle overwrite

I have noticed, from time to time, that some of our databases report in the RMAN output log as having two disk channels writing to the exact same backup piece handle. As expected, the associated backupset is corrupt.

I'm using the RMAN default format specifier for the disk device type along with an rman command like this:

backup incremental level 1 device type disk tag 'INCR' database to destination '/media/volDBbackup';

While I haven't figured out how to predictably get this to happen, when it does the log output looks something like this:

channel ORA_DISK_5: finished piece 1 at 01-AUG-2018 17:16:52piece handle=/media/volDBbackup/MYDB/backupset/2018_08_01/o1_mf_nnnd1_INCR__2959410745560_.bkp tag=INCR comment=NONEchannel ORA_DISK_5: backup set complete, elapsed time: 00:00:15channel ORA_DISK_6: finished piece 1 at 01-AUG-2018 17:16:52piece handle=/media/volDBbackup/MYDB/backupset/2018_08_01/o1_mf_nnnd1_INCR__2959410745560_.bkp tag=INCR comment=NONEchannel ORA_DISK_6: backup set complete, elapsed time: 00:00:15