We reproduced the problem in part with the latest backup from production.

Updating stats on all of the affected tables with FULLSCAN solved the problem.

We will run this tonight in production. Then I will let all of you know what happened.

Thanks so much for helpful hints.

John.