Suspect hit a bug, asked Oracle to confirm in the SR ==> seeing latch: row cache objects and enq: US - contention wait event during bad time for the problem sql when run at 2K/execution select SAMPLE_TIME, sql_id, SQL_PLAN_HASH_VALUE, event from dba_hist_active_sess_history where sample_time between to_date('2012-04-10 13:00', 'YYYY-MM-DD HH24:MI') and to_date('2012-04-10 14:00', 'YYYY-MM-DD HH24:MI') and sql_id='9x1h5r7vgd662'; output excerpt: 10-APR-12 01.30.14.337 PM 9x1h5r7vgd662 2573686592 latch: row cache objects 10-APR-12 01.30.14.337 PM 9x1h5r7vgd662 2573686592 latch: row cache objects 10-APR-12 01.31.25.555 PM 9x1h5r7vgd662 2573686592 enq: US - contention 10-APR-12 01.32.57.095 PM 9x1h5r7vgd662 2573686592 enq: US - contention 10-APR-12 01.35.49.826 PM 9x1h5r7vgd662 2573686592 enq: US - contention 10-APR-12 01.47.51.979 PM 9x1h5r7vgd662 2573686592 latch: row cache objects 10-APR-12 01.48.02.227 PM 9x1h5r7vgd662 2573686592 latch: row cache objects ... ==> Bug 7291739 -- Symptoms Waits for "latch: row cache objects" Waits for "enq: US - contention" -- Description High 'latch: row cache objects' contention may be seen on dc_rollback_segments together with high 'enq: US - contention' waits when using system managed undo with an auto tuned undo retention period. Not sure if the memory setting change especially from sga_target=0 to 12000M, which introduced ASMM could be the trigger of the bug. Denis -- //www.freelists.org/webpage/oracle-l