Explain Why Skipping an Earlier Stage Undermines a Later One

A stakeholder proposes: "Let's keep the canary release and governance review — those feel important — but skip the experiment tracking and registry work, since it seems like just bookkeeping and we're already disciplined about which model is deployed."

1. Walk through, concretely, what breaks in the canary-release stage if there's no reliable registry/tracking underneath it.
2. What breaks in the governance-review stage under the same condition?
3. Propose the minimum viable version of tracking/registry that would still support both of those later stages, if the team genuinely has limited time to invest.