Yeah an EC2 engineer switched over traffic to a backup network connection that had significantly less bandwidth, triggering cascading failures.