This makes a lot of sense. However, it makes me wonder how big is the new attack surface for remote upgrades/updates.
You need to implement a safe updater (with remote protocols) on VM level. And I guess you can never upgrade the VM itself, or if you can, it adds some extra complexity, or physical access.
There also need to be some kind of signature validation for every release, which means that device needs to perform some cryptographic operations and store at least tamper-proof public keys.