After a few months of hard work and extensive testing, we launched our new shiny feature. The first few hours after launch were calm and nobody could have guessed what was about to happen. Shortly after the full rollout, pages started to stall, users experimented with timeouts and system health dropped. At that point, we decided to turn on the kill switch and spend the next couple of weeks investigating our code’s efficiency. In this talk, I will cover what we did to find the bottleneck, analyze it, and finally, optimize our code to run 5 times faster.