Database hang caused by partitioning job(reference partitions) blocking application sessions

Environment:

Red Hat Enterprise Linux Server release 5.5

2 Node RAC

Issue:

Occasionally during the partitioning job runs(specifically during drop of reference partitions) the partitioning job goes through a deadlock with one of the application sessions, which gets cleared on timeout of 15minutes. Multiple application sessions are blocked during this period.

The partitioning job waits on a library cache lock and the application session waits on a row cache lock.

How they block each other is unclear.

A similar database with similar application logic but standalone and on a different OS does not have this issue.