CAs are a horrendous weakness, but are still a "pillar" weakness and not even a "base" problem. In other words the threat exists not because of a base technological capability or necessity, but because we _choose_ to stick our fingers in a blender and play with the switch. Like the liquidity deadlock persisting the post 1929 depression these things happen because humans, collectively are clinically insane in surrendering before systems (hubris, pride, social control). No one dares to override the system. In reality decisive emergency action can mitigate such sticky situations.

We also worry about existential threats from "class" points of weakness. Like a mathematical breakthrough that makes factoring large primes or mapping elliptic curves trivial. Again here, one can imagine that were it possible to entirely switch-off encryption and return to early 1990s operation, even a catastrophic class failure would not be that bad - we'd just have a massive cybercrime and infosec problem instead of a systemic failure.

Real dangers come from systems where logical regression is strictly not possible. A base weakness like a backdoor or fault in every ARM or Intel processor that could brick half the devices on Earth is a likely cause of widespread civic harm - because it could be unrecoverable, turning devices into e-waste overnight. Crowdstrike had a flavour of this because it reminded us that computers are actually pieces of hardware in data centres that can need manually rebooting or need human physical intervention at scale.