> Thanks Brandon, it looks like that confirmed the issue. > Thanks to everyone for their help. I really appreciate it Glad to see your problem has been solved. We recently had exactly the same case. A DBA from another team called and asked what could have caused a query to run with a good plan the first time and a re-run (by pressing /) uses a different, bad, plan in his 11gR2 database. I asked what column had "Y" in v$sql_shared_cursor for this SQL. It was use_feedback_stats for the bad cursor, he said. So we found bugs and notes on MOS and he ran alter session set "_optimizer_use_feedback"=false; before re-running the query. Problem solved. New technology may take some time to mature. For now, disabling it is a workaround for us. Yong Huang -- //www.freelists.org/webpage/oracle-l