We are experiencing a pga memory leak. So far I have not matched it up to a specific bug, and because of our environment we can't really work with support to narrow it down. I'll work on isolating the problem myself but....

I'm looking for a workaround. Typically we see the allocated PGA increase much faster than what is actually being used. I'm looking into seeing if we can workaround this with pga_aggregate_target & workarea_size_policy.

Is pga_aggregate_target a hard limit? IE, if I have a leaking PGA will it stop there? I'm sure it would depend on the leak, I'm just talking through it to see if the concept has merit.

As an aside, is there any way to free a processes unused, allocated PGA?

Thanks