Problems 'forcing' a better sql plan.

We are 19.8 2-node rac on exadata. The app is a file-processing application that has 24x7 activity. We struggle to keep stats up-to-date but lately they have not been stale often. We have moved away from plan baselines and have had better luck with sql profiles. This week a query was running quite poorly. We found the most efficient plan from awr. We gathered status on the underlying tables, created a sql_profile , stopped the app, flushed the plans for the this sql_id from the shared pool and restarted up the application. The