test: exercise durable review continuation

[khoi]

This temporary PR adds a clear arithmetic defect so SupaReview can show the complete review path in Actions. A follow-up commit will fix it and trigger the same PR conversation again to verify artifact-backed context.

Note

This PR exists only for end-to-end validation and will not merge.

[github-actions]

SupaReview

Findings

The new exported helper has a deterministic arithmetic error in its denominator, confirmed by both review passes. Although the PR says the defect is intentional for end-to-end validation, it is still the introduced correctness issue this review is meant to surface.

Findings

• [P2] Divide by the number of durations
  

  supareview/fixtures/durable-review.ts

  Line 3 in 22fe6cd

  return total / (durations.length - 1);

  
  averageDuration([100, 200]) returns 300 instead of 150 because the divisor subtracts one from the sample count. The same defect makes a one-element input return Infinity (and an empty input NaN), so every normal non-empty input produces an invalid average. Use durations.length and define the intended empty-input behavior before this exported fixture is consumed.

Review type: Full review

Reviewers: thermo-nuclear-review-subagent, thermo-nuclear-code-quality-review-subagent

[github-actions]

SupaReview

LGTM

Both incremental review passes found no qualifying issues. The change replaces the incorrect durations.length - 1 divisor with durations.length and explicitly handles the empty-input case, resolving the prior review finding.

Review type: Incremental review since 22fe6cd...7829731

Reviewers: thermo-nuclear-review-subagent, thermo-nuclear-code-quality-review-subagent

[khoi]

End-to-end test passed. The first run found the seeded bug. The next commit restored the same conversation, reviewed only the new commit range, and confirmed the fix.