We were just going to track and duplicate agents but would be much less code if we just had higher concurrency