You will probably not hit this bug unless you perform some media recovery in 18.104.22.168/AIX.
Ok. In case you hit
ORA-10878: parallel recovery slave died unexpectedly during a DUPLICATE or a RESTORE command, you can disable parallel media recovery with
The usual warning applies : do not use hidden parameter without guidance of Oracle Support. Open an SR if you hit this bug. Check for a patch on your plateform. Read notes 9728806.8 and 315631.1.
Note: for a RECOVER, the option RECOVER NOPARALLEL must be safer. Unfortunately there is no such thing like DUPLICATE NOPARALLEL
Update: This could also happened with standby, if you have stopped your standby site for a while and after restart you get ORA-10878 and ORA-00448 and evtl core dumps or internal errors, then stop dataguard (set dg_broker_start to false) and start the recovery manually with the noparallel option, until all logs are applied. Once this is done, you can restart dataguard, which will then in normal operation mode apply only one log at the time.
Diggout out from Helios’s Blog