Well, flushing the shared pool every time you execute the query still wouldn't work... because even the first time the query is compiled... it is compiled without looking at the bind values. Also, flushing the shared pool every time you execute a query is so sad I could cry.

As far as forcing the query plan you'd like with /*+ index */ hints (as well as nested-loops, hash, ordered, etc, etc) defeats the purpose of the CBO in the first place. One thing that the CBO has taught me is that regardless how smart I think I am at query optimization... once you're combining dozens of tables in all sorts of inner-views, unions, sorts, and merges... the CBO really can be smarter than me.

Also for my example, I could be running the same query on hundreds of different schemata, each of which have the same table structure, but contain a different demographic of data. The CBO is going to pick the proper execution path for the schema it is being executed in.

Anyway... I'm familiar with all of those issues, and I'm really going to end up just altering the application code in which this query is executing, so that I can parameterize the sql text itself (so I don't have to bind these literals). I was just curious if you knew some cool trick that I didn't. :-D

Thanks, though.