Size a Canary Rollout for a Weekly-Cadence, Financially-Impactful Model

The churn model scores subscribers weekly, and the retention team acts on the top 5% with a real-dollar discount offer. A teammate proposes skipping shadow mode and canary entirely, arguing "we already passed the offline champion/challenger gate, that should be enough."

1. Explain concretely what the offline gate does and does not tell you that a canary would catch.
2. Propose a specific canary design (population %, duration, guardrail metrics) for this model, and justify each choice given its weekly cadence and financial impact.
3. What automatic rollback trigger would you define, and why is "flagged-list overlap week over week" a meaningful guardrail here specifically?