Baking the VM images trades a lot of flexibility for some latency, and it doesn't apply to container-as-a-service infrastructure where you don't control the VM.

Baking the VM images trades a lot of flexibility for some latency, and it doesn't apply to container-as-a-service infrastructure where you don't control the VM. My point is you can get most of the way there with slim/distroless without making the tradeoffs of Alpine, and that's probably enough for most use cases.