I'm curious about how you turn the Docker image into a root filesystem for the micro-VM. If you're willing to share more about this, are you using an existing tool such as LinuxKit or Packer, or did you write your own?
We built something into our registry that squashes layers of an image into a compressed rootfs archive. Edge nodes map an image back to one of these files when launching an app. This cut launch time for large images in remote regions from tens of seconds to a few hundred ms. Much of that infrastructure is actually running on fly itself!