Between approximately 4:00 AM and 10:00 AM US Pacific time on 15 December, customers on our US cloud site may have experienced an elevated level of error responses from our various Named User API endpoints. As such, unless those requests were successfully retried, this may have prevented our systems from getting updated information about devices and sending messages to them if they were targeted by named user or named user tags. The issue was resolved by adding more compute resources to the service. Later on, our Backend Engineering team discovered that a significantly large amount of CPU on named user API instances was being used up by unnecessary logging infrastructure, which they then removed. This will prevent future recurrences of the same service defect. Engineering is also planning to improve monitoring and alerting for this API so they will become aware of similar problems and resolve them more quickly in the future.