We have identified a router that was causing brief connectivity issues (two ten minute periods) for a very small set of instances in a single Availability Zone in the US-EAST-1 Region. We have removed the problematic router and the service is operating normally.