Fix Failing E2E Tests

Fix failing E2E tests by modifying test implementations. This command NEVER modifies E2E_TESTS.md spec files — the spec is the source of truth.

Additional instructions from the user: "$ARGUMENTS". Ignore if empty.

Phase 1: Identify Failures

1. Run the E2E test suite to capture current failures.
   • If the user specified a particular test or suite in $ARGUMENTS, run only that.
   • Otherwise, run the full E2E suite.
2. For each failure, capture:
   • Test file and test name
   • Error message and stack trace
   • Expected vs actual values

Phase 2: Diagnose

For each failing test:

1. Read the failing test file to understand the test implementation.

2. Read the corresponding E2E_TESTS.md spec to understand the intended behavior.

3. Classify the failure:

   Test bug: The test implementation is wrong (bad assertion, incorrect setup, missing cleanup).

   • Fix: Correct the test to match the spec.

   Stale test: The code under test changed but the test wasn't updated.

   • Fix: Update the test to match current code behavior AND verify the spec still matches.
   • If the spec is outdated, inform the user but do NOT modify the spec.

   Environment issue: Missing dependency, port conflict, timing issue.

   • Fix: Make the test more resilient (add retries, increase timeouts, fix resource cleanup).

   Flaky test: Passes sometimes, fails sometimes.

   • Fix: Identify the source of non-determinism (timing, resource contention, ordering) and eliminate it.

Phase 3: Fix

Apply fixes following these rules:

• Never modify E2E_TESTS.md: If the spec seems wrong, report it to the user but don't change it.
• Follow existing patterns: Match the code style and patterns of adjacent passing tests.
• Maintain isolation: Fixes must not introduce shared state or ordering dependencies.
• Preserve assertions: Fix HOW a test runs, not WHAT it asserts (unless the assertion contradicts the spec).
• One fix at a time: Fix one test, verify it passes, then move to the next.
• Red/green TDD: When fixing a test, first confirm the failure is reproducible (red). Apply the minimal fix. Re-run to confirm it passes (green). Then run the full suite to check for regressions. When manual testing or debugging reveals an issue not covered by existing tests, add a spec entry and generate a test that fails (red), then fix the underlying code or test (green).

Common fixes:

Missing cleanup:

// Add proper afterEach if missing
afterEach:
  changeDirectory(originalCwd)
  env.PATH = originalPath
  safeCleanup(testDir)

Timing issues:

// Add polling with timeout instead of fixed waits
function waitFor(condition, timeout=30000, interval=500):
  start = now()
  while now() - start < timeout:
    if condition():
      return
    wait(interval)
  throw TimeoutError("condition not met within {timeout}ms")

Resource conflicts:

// Use unique resources per test
testDir = createTempDirectory("unique-prefix-")  // Not a fixed path

Phase 4: Verify

1. Run the previously failing tests to confirm they now pass.
2. Run the full E2E suite to confirm no regressions.
3. Run the project's formatter on all modified files.
4. Present a summary:
   • Tests fixed
   • Root cause for each fix
   • Any spec issues found (for user to address separately)