A release decision should preserve the comparison that supports it. More total passes can coexist with a new critical regression. Keep the baseline when mandatory gates fail or the agreed comparison introduces a new failure, and record which evidence would justify trying again.
Return promote only if case.release.candidate_safety is true, candidate_passed is at least baseline_passed and new_failures is empty. Otherwise return retain-baseline. Improved aggregate counts do not override a new failure.
The notebook executes your Python against an original local mock. Use the supplied case rather than copying visible values. The hidden checks change identities, revisions and boundary inputs; no provider endpoint or external tool is contacted.
An example implementation of the function contract.