“Timelocks and multisig have classes of uses beyond lightning so you’re right, it would have been silly to reject them because we didn’t know of the potential of lightning. We already knew other ways they could be used. For example, shared custody.”

You misunderstand my analogy here. It wasn’t that we didn’t understand it’s value because of its, it’s that one didn’t have to defend multisig as a tool by predicting everything that could be built with it to know it was safe. If that was a requirement, then we could never make any changes at all. 

“Imagine you are a systems designer for a nuclear reactor…”

Dude I’ve done dozens of shows on this, you don’t need to argue that this is foundational, nuclear grade software, that this should be treated with utmost care, that upgrades should be extremely limited in scope and well understood. If you are making that argument then we aren’t actually on topic.