[anth-volk]

Fixes #4006

Summary

The simulation service owns simulations, reports, report runs and their outputs in its own simulation schema. A later autogenerated v2 revision (the next PR in this stack) drops the Public API's copies of those 15 tables, their 8 PostgreSQL enum types and the retired country and label columns of the simulation and report associations. The release workflow upgrades the schema before the new Cloud Run revision serves traffic, so the release serving traffic while that revision runs must never touch them. This PR produces that release. It adds no Alembic revision, and the tables and columns stay in the database.

• Catalog publication no longer counts rows in simulations, reports and report_runs to prove it changed none; it counts only dataset_versions. The catalog writer's declared SELECT on those three tables is removed from the privilege manifest. The deploy step never revokes, so the existing grants stay (reported as undeclared in the job summary) until the next PR drops the tables.
• The association models no longer declare country and label. A new module, models/retired_tables.py, appends the two columns to the two tables after their mappers exist, so no mapper selects, inserts or updates them.
• The simulation-domain models are removed: models/reports.py, models/simulations.py, HouseholdJob, the unused report_runs module and their tests, and every relationship the kept models (users, policies, dynamics, households, datasets, regions, model versions) had to them. Deleting a user through the ORM no longer selects from reports; the foreign keys and their ON DELETE actions stay in the schema.
• retired_tables.py declares the 15 tables as plain SQLAlchemy tables that no model maps. Their compiled PostgreSQL DDL (columns, types, defaults, keys, checks, indexes, enum names and values) is identical to the removed models'. The v2 metadata therefore still matches the schema at head (alembic check reports no operations), and deleting that module is what lets autogeneration produce the next PR's revision.
• The household migration qualification counts simulations, reports and household_jobs through table() constructs instead of the models. It still skips tables a target does not have.
• Tests of the retired tables' design rules are removed, and the Stage 9, Stage 12A and testing documents are updated.

Migration scope: no route group, workflow or user-visible API contract changes. Nothing moves off the Flask/API v1 path.

Inline suppressions added: one # noqa: E402 on the retired_tables import in policyengine_api/data/v2/models/__init__.py. The package must set the metadata naming convention before any table is declared, so every import in that file follows the assignment and carries the same suppression; the next PR removes this import.

Environment variables

None. The changed code reads and declares no environment variable. The changed privilege declarations reach staging and production through the existing scripts/reconcile_v2_privileges.py apply step in .github/workflows/seed-v2-database.yml, whose variables and wiring this PR does not change.

Merge requirements

• CI and an approving review. No other PR or operator step blocks this one.
• It must be deployed to staging and production before the next PR in the stack (which drops the tables) merges.

Testing

• uv run pytest tests/unit/v2 tests/contract -q: 1036 passed. The rest of tests/unit and tests/to_refactor: 2331 passed.
• New tests record every SQL statement the ORM sends and fail if any names a retired table, or names country or label on the association tables: association insert, read, update and delete, and user deletion. They fail on master (the association insert lists country and label; user deletion selects from reports).
• Model tests: no model maps a retired table, the models package exports none of the removed names, and no kept model has a relationship to a retired table.
• Against disposable PostgreSQL: alembic -c alembic-v2.ini check reports no new operations; tests/integration/test_alembic_v2_lifecycle.py, test_v2_catalog_publication.py (including a new test that publishes as the catalog writer after its access to the three tables is revoked), test_v2_user_identity_persistence.py, test_v2_privilege_reconciliation.py, test_v2_stage12_comparison_persistence.py, test_v2_household_persistence.py, test_v2_policy_persistence.py, plus the cross-database dual-write tests against disposable MySQL: all passed.
• mypy, ruff format --check, ruff check, scripts/run_quality_guards.py and scripts/export_migration_contracts.py (no changes): all passed.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 98.27586% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 87.72%. Comparing base (ee46c7a) to head (943601c).

Files with missing lines	Patch %	Lines
policyengine_api/data/v2/catalog/publication.py	75.00%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##           master    #4007       +/-   ##
===========================================
+ Coverage   50.12%   87.72%   +37.59%     
===========================================
  Files         243      257       +14     
  Lines       14726    15768     +1042     
  Branches     2343     2616      +273     
===========================================
+ Hits         7382    13833     +6451     
+ Misses       6721     1265     -5456     
- Partials      623      670       +47     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.