Between 7:59 AM and 10:05 AM PST, we experienced increased error rates with the CloudWatch GetMetricStatistics and GetMetricData APIs when requesting metrics older than 3 hours in the US-EAST-1 Region. CloudWatch alarms may have transitioned into “INSUFFICIENT_DATA” state if the period of the alarm was longer than 3 hours. Querying metrics less than 3 hours as well as publishing metrics were unimpacted. The issue has been resolved and the service is operating normally.