Establish a reproducible app baseline for the three-commons dogfood

Nobody has claimed this yet.

A dogfood result is hard to interpret when deployed behavior differs from the checkout or browser tests interfere with the live application. The app repository's September 4 handoff records hosted test interference, incomplete phase admission and a newest-change attribution gap. Reverify those findings before treating them as current defects.

## Work

- Establish an isolated, reproducible database and fixture setup for browser acceptance tests.
- Record the tested application commit, migrations and deployed function versions; verify the live agent's meaning-confirmation path.
- Reproduce and resolve any remaining inability to identify the proposer and ratifier of the newest accepted change.
- Run the admitted acceptance profile plus affected new slices, distinguishing product failures, fixture failures and deferred behavior.
- Replace stale opening status claims in the app handoff with one current verified summary.

## Acceptance criteria

A repeatable test command targets an isolated environment. The setup → contribution → interpretation correction → decision → return journey works across two member sessions; access boundaries between commons are checked. Current-change attribution is reachable. Results identify the exact tested versions and unresolved failures.

## Expected benefit and check

David and Rufus can attribute dogfood friction to the product rather than an unknown deployment or test outage. Demonstrate the journey before relying on the dogfood measurements.

Implementation belongs in the reason-commons app checkout; this issue coordinates the work in the shared project repository. This is a bounded reliability gate, not a request to complete every future redesign phase.

Start in the reason-commons app checkout by reading the September 4 app handoff and its recorded findings on test interference, phase admission, and change attribution. Run the admitted acceptance profile and affected new slices in an isolated environment across two member sessions. Done means the journey works, access boundaries are checked, versions and unresolved failures are recorded, and the handoff has one current verified summary.

Written by the indexing model from the issue text.