Deployment_platform is relatively expensive and calling this after the
fact means that this may not be the cluster that was deployed to.

Correspondingly reduce the leeway given in the related N+1 spec