This finally got committed, after a lot of performance and testing rework, and discovering plenty of opportunity for future performance, scalability and testing improvements. Thanks to reviewers for performance push back, leading especially to a better ArrayDeque.