GitHub has been enhancing its application performance by incorporating a feature called rack.after_reply into its Unicorn-powered web servers, resulting in significant improvements in response times. Previously, the process of sending telemetry metrics during requests was causing delays, with up to 65ms added per request. By implementing rack.after_reply, GitHub can now batch and flush telemetry data only after users receive a response, reducing the response time by 30ms at the median (p50) and more than 50ms at the 99th percentile (p99). Although this approach initially posed challenges, such as blocking new HTTP requests until the rack.after_reply code execution completed, GitHub mitigated these issues with a timeout feature. This development not only enhances performance but also contributes a valuable tool to the Unicorn web server, with potential inclusion in the official Rack specification.