Looking for suggestions - how find cause of sql invalidations

TL;DR: How to find cause of high sql invalidations when nothing obvious is happening (such as stats or table ddl).

We're having some library cache/shared pool thrashing and I ran an AWR for a 2 hour period this morning.

In the library cache activity, for SQL AREA, I see: sql_id = 49s16x6k5ubwc
Invalidations_Total = 2065
Invalidations_Delta = 1287

This SQL_ID is executed many, many times every hour and I'm trying to figure out why it's getting invalidated.

So, far I've checked v$active_session_history and dba_hist_active_sess_history to get the full history of the sessions that have executed that SQL_ID.

Nothing in those sessions appear to be doing any type of DDL (stats or anything like that) so it appears to be something outside those sessions causing the cursor invalidations.

There is one table involved in the query - and it has 1444 partitions.

I checked LAST_DDL_TIME from dba_objects and I don't see any DDL timestamps that would be related.

I've checked last_analyzed from DBA_TAB_PARTITIONS and that the table hasn't been analyzed during the period of invalidations.

Much of the SQL being executed against this table is "EXECUTE IMMEDIATE" from within pl/sql.

One of the code blocks does have an IF statement that would do a "LOCK TABLE" if the condition is met (I can't tell if that condition is being met however). Would a LOCK TABLE cause invalidations?

Any suggestions on how to crack this particular egg?

Chris

--
http://www.freelists.org/webpage/oracle-l