Some parts of my post were just wrong, they're edited now. But other parts use the unspoken assumption that there's such a thing as "standard integers" (or, equivalently, there's such a thing as "Turing machines") and the axioms of PA are true statements about that thing. That seems to imply omega-consistency, but the whole argument is so informal that I can't tell for sure. It could be formalized somehow, I guess, but that was not the intent.

In the words of Liron Shapira, I'm talking about Turing machines as "their own meta-level thing", so statements about their halting or non-halting are to be interpreted as "facts of the matter" outside any formal system. The "standard integers" exist in the same limbo. That's where the handwavy reasoning about SSS...S0 comes from.

Yeah I know that's weird.