AWS lambda runtime has a custom http protocol for invocations, that part should be rather easy.

Then there is a configuration for the runtime, sort of infrastructure dependencies.

Next the resource permissions (storage, databases, ...).

Probably a lot more, but just the idea that you could then easily move a Lambda function over to a #DVM makes me think that cloud infrastructure could become a market for smaller compute miners again.