The root cause of this incident was an outage at the AWS level. Because the outage was region-wide, no remediation was possible on our end. We needed to wait for AWS to bring the region back online. Because this only affected traffic for tests running through the São Paulo, Brazil geolocation, the impact was minimal. After this incident, we considered the idea of rolling traffic over to a different geolocation in this situation, but determined that it is best to let the tests fail and make it clear that the geolocation is having issues. We believe that’s a more useful course of action than redirecting traffic in an unexpected way.