We are migrating sandbox infrastructure. No downtime should happen except at the point of cutover when in-progress requests will fail. All requests should succeed upon retry.