I really like the vision of sprites as: cost practically nothing while asleep. However, just like paying NVMe storage rates for sleeping sprites goes against that vision, so does paying $1/mo for unused concurrency.
Please consider adding an option for the $0/mo pay as you go plan to add concurrency on a per-second basis. For example, perhaps apply a multiple, like 4x, similar to sprite vs. Fly machine CPU and RAM pricing, and charge an additional $0.005/hr per running sprite above the baseline concurrency allowance?