Cloudflare has made an optimization to its Workers runtime that reduces the amount of memory usage for Unbound Workers. The change allows the system to mark application code as "idle" as soon as response headers are returned, reducing the time a Worker is considered "in use" by an average of 70%. This change also applies to WebSocket proxying, reading from the cache, and streaming from KV. As a result, users may see a drop in their billing for these services.