Codify Deterministic Test Fixtures (codex)

[numnx]

🧩 Task: Codify Deterministic Test Fixtures

Automated task PR opened by Code UX for T02.

📋 Summary

Sprint: Sprint 144 · QS: Coverage Observability and CI Reliability Audit · Task: T02

Create a repeatable coverage, observability, security, and CI reliability quicksprint by adding enforceable tests, CI/Playwright hardening, structured logging guarantees, and aligned operational documentation without lowering coverage thresholds.

📝 Full Task Prompt

## Objective
Close the current verification gap where deterministic test assumptions live partly in convention and can regress silently. The desired guarantee is that Vitest setup keeps isolated app state, UTC/C locale defaults, explicit fake-timer use, and safe cleanup behavior across backend tests.

## Scope
- Edit `tests/setup/runtime-warning-filter.ts` only if a helper or exported assertion metadata is needed
- Create or edit `tests/backend/scripts/test-determinism-guardrails.test.ts`
- Verify `vitest.config.ts`
- Verify representative tests under `tests/backend/shared/polling/wait-until.test.ts`, `tests/backend/repositories/sqlite-connection.test.ts`, and `tests/backend/services/provider-concurrency-service.test.ts`
- Relevant symbols: `withIsolatedTestHome`, `VITEST_IN_MEMORY_DB`, `HOME`, `USERPROFILE`, `XDG_CONFIG_HOME`, `XDG_STATE_HOME`, `XDG_CACHE_HOME`, `vi.useFakeTimers`, `vi.useRealTimers`

## Implementation Requirements
1. Add tests that exercise `withIsolatedTestHome` and assert it sets and restores `HOME`, `USERPROFILE`, `XDG_CONFIG_HOME`, `XDG_STATE_HOME`, and `XDG_CACHE_HOME` for both successful and throwing callbacks.
2. Add a repository-level guard test that reads `vitest.config.ts` and asserts deterministic locale/timezone settings are present and that `environment` remains `node` by default.
3. Add a guard test that scans backend test files for `vi.useFakeTimers()` and requires nearby `vi.useRealTimers()` or `afterEach` cleanup in the same file, with an allowlist only if a file has a documented reason.
4. Add a guard test that verifies tests using filesystem app state or SQLite helpers either rely on `VITEST_IN_MEMORY_DB=true` or use `withIsolatedTestHome`/temporary directories instead of a real user home.
5. Keep failure output actionable by listing the exact offending file paths and the missing cleanup category.

## Constraints
- Do not add brittle line-number assertions.
- Do not require fake timers globally; the intended guarantee is explicit fake-timer setup and cleanup per test file.
- Do not invoke real Docker, Git remotes, provider CLIs, or persistent SQLite files.
- Preserve current runtime warning filtering behavior for ordinary test runs.

## Verification
- Run `pnpm run test:backend -- tests/backend/scripts/test-determinism-guardrails.test.ts` or the exact focused file created.
- Run `pnpm run test:backend -- tests/backend/shared/polling/wait-until.test.ts tests/backend/repositories/sqlite-connection.test.ts tests/backend/services/provider-concurrency-service.test.ts`.
- Run `pnpm run lint` and confirm the guardrail tests have meaningful failure messages.

✅ QA Review

⏳ QA review runs after this PR is opened. Check the task card in the dashboard for the outcome once it lands.

🤖 Model & Provider

Provider	Codex CLI
Model	gpt-5.5

⏱️ Timing

Started	2026-07-06 20:09:41 UTC
Finished	2026-07-06 20:15:34 UTC
Duration	5m 53s

📊 Token Usage

Metric	Value
Input tokens	112,497
Cached input tokens	1,286,784
Output tokens	10,156
Total tokens	1,409,437
Invocations	1
Tool calls	41
Active time	5m 14s

Included usage estimate (subscription/local login): $1.51

🌿 Branch Info

Base: feature/CODUX-144-qs-coverage-observability-and-ci-reliability-audit-4
Head: task/feature-codux-144-qs-coverage-observability-and--t02-codex-mr9nlx4a

---

🤖 Generated by Code UX