A small number of customers were affected by a routing error within our web load balancers. This affected the routing of their traffic towards a single unstable Gateway API. This would have affected some other dependent services. The routing error has been corrected and we are working to implement further error prevention measures.