Earlier quoted context omitted.
Do you have a link to that? Because I know a guy who really needs to counterpoint it. High-security engineer, Clive Robinson, always said security is about physics if you look at it down to the hardware. The physics try to connect things in ways you didn't see coming. That allows unauthorized communications. The physics also try to corrupt the operation of your chips. That compromises computational security mechanism…
Security != encryption in every case. What you're describing is actually also what makes encryption stronger/easier than decryption: A priori there's only 1 correct plaintext, while there are limitless chipertexts of any given plain text (assuming arbitrary IV lengths and key). You can't change that and this is basically what makes encryption so much stronger than decryption.
Add it all up to say that, outside a few products, your security mechanisms from CPU go crypto arent secure. Physics and intrinsic complexity work together to ensure this. Systems fighting all of it have less features, are heavy, more manual steps, less battery life, and cost several times more. Economics takes over there where physics leaves off.
"A priori there's only 1 correct plaintext, while there are limitless chipertexts of any given plain text (assuming arbitrary IV lengths and key)."
A priori there's electrical signals going through analog and digital circuitry that implements a form of it with malicious hardware, software, or networks connected to it. There's tons of ways to intercept or leak those secrets. These are not in the formal model of crypto. Once included, the picture changes considerably and leans my way.