Audit cloud test coverage and reduce over-mocking

Problem

The current Cloud test suite leaned heavily on monkeypatched unit tests for view logic and helper functions. That kept tests fast, but it also left a blind spot: fake querysets and patched render/context helpers could validate the shape of a function call while missing real Django ORM, queryset, URL, and template failures. The /piedpiper/ crash was a concrete example: a real Dashboard queryset raised on order_by("-updated_at"), but the existing test passed because it used a fake queryset that accepted any field name.

Context

• Recent bug:
• apps/cloud/apps/chat/dashboard_cards.py ordered Dashboard by -updated_at, but apps/cloud/apps/projects/models.py::Dashboard only has last_synced.
• The failure surfaced only when loading the real org home/chat page.
• Existing Cloud tests were concentrated in monkeypatched request/view tests:
• tests/cloud/test_org_chat_home.py
• tests/cloud/test_chat_persistence.py
• tests/cloud/test_dashboard_embed.py
• Constraints:
• Keep the suite fast enough for regular local use and PR gating.
• Prefer the smallest realistic harness that exercises real Django ORM and template rendering on critical paths.
• Avoid jumping straight to a broad browser suite for server-rendered page-load failures.

Possible Solutions

• Keep patching individual regressions as they appear.
• Fastest short-term path.
• Not recommended because it preserves the same structural blind spot.
• Add a small ORM-backed/request-backed smoke layer for critical Cloud routes, while keeping narrow monkeypatched unit tests for pure branching logic.
• Recommended
• Exercises real models, querysets, view wiring, and templates for the highest-risk server-rendered routes without adding browser-test overhead.
• Build broad browser-level Playwright coverage immediately.
• Valuable later for JS and htmx flows.
• Too heavy as the first response to this failure class.

Plan

1. Audit the existing tests/cloud coverage and identify tests that stub out the framework layers most likely to fail in production.
2. Add DB-backed request smoke coverage for high-value server-rendered routes: org home, project home, and dashboard view.
3. Fix the concrete queryset regression by ordering recent dashboards on the real last_synced field.
4. Remove or narrow redundant tautological tests that only assert mocked ORM strings instead of real behavior.
5. Document the recommended Cloud testing split so future tests do not drift back toward over-mocking.

Implementation Progress

• Audited the current Cloud tests and confirmed there was no DB-backed request/template smoke coverage in tests/cloud.
• Fixed apps/cloud/apps/chat/dashboard_cards.py to order recent dashboards by -last_synced, which matches the real Dashboard model.
• Added tests/cloud/test_cloud_route_smoke.py:
• boots a minimal Django schema locally with migrations
• resets the DB between tests with flush
• exercises real ORM + real template rendering for:
  • org home/chat page
  • project home
  • dashboard view
• adds an explicit regression assertion that get_recent_dashboards() orders by last_synced descending
• Removed the tautological fake-queryset ordering test from tests/cloud/test_org_chat_home.py and kept the remaining narrow monkeypatched tests focused on branching/render-selection behavior.
• Updated apps/cloud/README.md with a Cloud test split:
• monkeypatched tests for pure logic
• ORM/request smoke tests for critical server-rendered routes
• browser tests only where JS/htmx behavior actually matters

QA Exploration

• N/A for this task. Coverage added at the Django ORM/request/template layer.
• [x] QA exploration completed (or N/A for non-UI tasks)

Review Feedback

• Initial review flagged the old fake-queryset ordering test as redundant and asked for an explicit ORM-backed ordering assertion.
• Addressed by deleting the tautological ordering test and adding a real last_synced ordering regression test in tests/cloud/test_cloud_route_smoke.py.
• [x] Review cleared