Something on my #AWSWishList: brownout as a service. I'd like to be able to inactivate a thing without otherwise hurting it, across many services. Then emit a Health event for "inactivated thing <service>.<instance> tried to do something". Then I could automate lengthening brownouts to verify it's unused before reporting it's OK to terminate. But it'd need tiny changes to many resource types in many services, so I suppose unlikely.