Due to a configuration error, our network DDoS prevention system was unintentionally throttling traffic to our backend DNS servers. Because this occurred at our network edge, our monitoring showed the DNS service itself was performing normally. We worked closely with Cloudflare, who provides geographic redundancy for our DNS services, to determine that requests were terminating inside our network, allowing us to pinpoint the problem. We've resolved the configuration errors within the involved systems, and we do not anticipate any additional issues going forward.