From 0:28 to 0:58 (UTC), our system experienced a significant degrade in performance, causing ~50% of user requests to take more than 5 seconds to respond.
The root cause was due to a memory issue from a recent coding change. We've immediately released a fix for this.