Speed up slow tests without losing coverage

Why use this

1. An engineering team may want faster tests without losing confidence in what they check. A refactor can reduce runtime by accidentally skipping cases, weakening assertions or changing the environment in ways that make the comparison unfair.

2. This job could prepare a performance-change proposal with explicit preservation checks and review evidence. Keeping speed, test inventory and behavioral coverage separate helps reviewers avoid treating one favorable metric as proof that the suite remains equivalent.

Why an agent / simpler alternative

Useful judgment

Identify likely bottlenecks, propose bounded refactors and explain what evidence would be needed to assess preserved behavior.

Simpler alternative

Profile and adjust a small slow test manually when the cause is obvious.

Unsuitable when

The workflow would remove checks, weaken assertions or claim equivalent coverage from counts alone.

Proposed inputs

1. Name

   Test-suite evidence

   Owner

   Maintainer

   Minimum scope

   Approved timings, test inventory and relevant code summaries

   Quality check

   Record environment and baseline version

2. Name

   Preservation criteria

   Owner

   Quality owner

   Minimum scope

   Required assertions, coverage expectations and permitted refactors

   Quality check

   Separate speed goals from correctness requirements

Access and setup — proposed or unknown

Proposed minimum

1. Review fictional timing/coverage summaries and a proposed diff; no profiling, test execution or repository edits.

Compatibility unknowns

1. Current test tooling, coverage semantics, timing stability and runtime compatibility are unknown.

External actions requiring approval

1. Code changes, dependency updates, compute runs and pull-request creation requires scoped authorization and review.

Expected deliverable and review checks

Proposed artifact

Refactor-plan candidate with baseline comparison, preservation checks and reasons to reject or hold a change.

Critical checks

1. Use comparable environments; inspect skipped tests and assertions; distinguish covered lines from adequate behavior testing; retain failed validations.

Missing output evidence

This overlay adds no new output inspection, execution or validation; consult the referenced canonical source outcome limits.

Failure signals and proposed responses

1. Signal

   A test disappears while total count stays similar

   Why it matters

   An added trivial case can mask lost coverage

   Proposed response

   Compare test identities and assertions, not totals alone.

2. Signal

   A warm-cache run is compared with a cold baseline

   Why it matters

   The apparent improvement may not reflect the change

   Proposed response

   Define consistent measurement conditions.

Effort and cost — unknown

Setup work

1. Define baseline and preservation evidence.

Ongoing review

1. Maintainers review actual results and the diff before accepting changes.

Unknowns

1. Timing variance, coverage adequacy, refactoring effort and compute costs are unknown.

First check — proposed, unexecuted

Status

proposed_unexecuted

Synthetic fixture

Invent before/after summaries with a skipped case, unchanged count and different cache conditions; review the proposed evidence only.

Acceptance checks

1. Flag both comparability failures; infer no speedup or preserved correctness; run no tests.

Stop conditions

1. Stop if baseline comparability or essential assertion preservation is unresolved.

Limits

1. A comparison plan cannot establish faster execution, equivalent coverage or unchanged software behavior.