IIRC in the wake of Spectre, Google had a security blog that noted that even at Google with all the layers of software involved they felt that there really was no way to ensure that something in those layers isn't running something unexpected / doing something it shouldn't.

IIRC The general topic was speculative execution and such security, but I think it also made it clear that more generally all the chips, software involved ... you just can't know.

To be clear, I don't say that to dispute anything you're saying, quite the opposite.