Before every deployment, explicitly identify the single most probable failure mode for that specific change and pre-agree on the immediate indicator of this failure, ensuring the monitoring team knows exactly what signal to prioritize.