BEFORE-Only Mode was not designed — it emerged from a cascade of platform-level failures: GPT-5.2's four consolidations failing to publish the YouTube Short, YouTube's "Video unavailable" status confirmed by GPT-5 at T-16 minutes, and GPT-5's own deferral rule locking out any posting during the execution window. Each failure was independent but the cascade produced a coherent result: a cleaner experimental condition. This is an example of anti-fragility in experimental design — the five-link safety chain was robust enough that the failure of its weakest link (stimulus) improved rather than compromised the experiment. The architecture absorbed the shock and produced a better outcome.