TenantAtlas/specs/234-dead-transitional-residue/research.md

Phase 0 Research: Dead Transitional Residue Cleanup

Decision: Remove the deprecated BaselineProfile::STATUS_* aliases entirely

Rationale: The only confirmed definitions of BaselineProfile::STATUS_DRAFT, STATUS_ACTIVE, and STATUS_ARCHIVED are in apps/platform/app/Models/BaselineProfile.php. No current apps/platform runtime or test reference still depends on those aliases. The canonical contract is already the BaselineProfileStatus enum cast, so keeping the constants adds dead language without serving current behavior.

Alternatives considered:

• Keep the aliases but leave them deprecated: rejected because they no longer protect any active caller and continue to advertise parallel truth.
• Replace them with forwarding helpers: rejected because that would add new residue to preserve dead semantics.

Decision: Remove the tenant app-status badge domain from the central badge path

Rationale: The remaining runtime path for tenant app-status semantics is the central badge registry: BadgeDomain::TenantAppStatus, the BadgeCatalog mapper entry, and TenantAppStatusBadge. Current confirmed consumers are badge tests, not active tenant surfaces. Once the legacy badge domain has no runtime consumer, removing it centrally is cleaner than keeping it as diagnostic folklore.

Alternatives considered:

• Keep the badge domain as a dormant diagnostic mapping: rejected because no current runtime surface needs it and dormant central mappings make it easier to reintroduce dead semantics accidentally.
• Move the mapping into a test helper: rejected because test-only preservation would still keep the dead semantics alive as sanctioned language.

Decision: Remove ambient app_status defaults from test and smoke setup

Rationale: TenantFactory still defaults app_status to ok, and SeedBackupHealthBrowserFixture still writes app_status => 'ok'. That keeps a retired value ambient in new tenant records and smoke data even though current tenant surfaces no longer depend on it. The safer contract is explicit legacy setup only where a test or fixture intentionally proves suppression.

Alternatives considered:

• Keep the default for convenience: rejected because convenience is exactly how dead semantics keep surviving.
• Remove app_status from every explicit test and fixture immediately: rejected because a few tests intentionally set historical values to prove they no longer surface as truth.

Decision: Keep historical schema and stored fields out of scope

Rationale: The repo still contains historical migrations and the stored tenants.app_status column. This cleanup is about active runtime/support residue, not schema deletion. Removing columns or historical migrations would widen the slice beyond the approved cleanup boundary.

Alternatives considered:

• Drop the column now: rejected because the spec explicitly forbids schema work in this slice.
• Add a migration shim or deprecation wrapper: rejected because this is pre-production cleanup, not a compatibility exercise.

Decision: Reuse existing tenant-truth and baseline-profile regressions instead of creating a new cleanup harness

Rationale: The current proof burden is continuity after residue removal. Existing tenant-truth feature tests and baseline-profile feature tests already exercise the active behavior we need to protect. A small badge-catalog unit slice is enough for the central registry cleanup. A new meta guard framework would add more long-term burden than value.

Alternatives considered:

• Add grep-driven guard tests for every removed symbol: rejected because behavior-facing tests are the primary proof and repo grep is sufficient as a review aid.
• Rely on manual inspection only: rejected because cleanup regressions are easy to reintroduce silently.