how is this achieving scale-up and scale-to-zero ? from my (rudimentary) investigation only knative k8s had scale-to-zero implemented well enough.

and if you dont have scale-to-zero, you cant claim a vercel alternative.