The strongest objection
A theory becomes difficult to evaluate if every success is called temporary and every failure is treated as confirmation. If no conceivable evidence could reduce concern, the position risks becoming insulated from reality. W360's prevention argument should therefore make its assumptions and possible revisions explicit.
Evidence that would matter
Evidence of robust control at relevant capabilities would matter, especially independently reproduced results showing that safeguards withstand realistic attempts to defeat them. The evaluation would need to cover the system's operational access and combinations with other tools, and explain how the assurance carries over to deployment.
Evidence that dangerous capabilities face durable physical or economic barriers could weaken expectations of rapid escalation. Evidence of stable institutions that preserve human authority despite extensive automation could weaken the claim that loss of economic necessity inevitably undermines political agency.
Policy evidence matters too. If a proposed restriction proves unenforceable, drives development beyond observation, or creates greater harms than a workable alternative, its design should change. A strong diagnosis does not make every proposed remedy effective.
What would not settle it alone
A period without a catastrophe, a reassuring demonstration, or agreement among prominent people would not by itself establish long-term safety. Conversely, one failure would not prove that every possible safeguard is futile. The hinge is whether accumulating evidence supports the causal claims at the capabilities and deployment conditions being discussed. This site's arguments and proposals should be revised as that evidence changes.