[nwparker]

	Files	Added	Deleted	Net
Test	421	$\color{#1a7f37}{\Huge{\mathbf{+}}}$​127	$\color{#cf222e}{\Huge{\mathbf{−}}}$​20422	$\color{#cf222e}{\Huge{\mathbf{−}}}$​20295
Prod	17	$\color{#1a7f37}{\Huge{\mathbf{+}}}$​59	$\color{#cf222e}{\Huge{\mathbf{−}}}$​1173	$\color{#cf222e}{\Huge{\mathbf{−}}}$​1114

ELI5

PR checks spend time running source inventories and repeated comparisons that add little confidence. Remove those tests so runners spend more of their time checking real behavior.

What Changed

Remove 132 test files, retain five meaningful contracts in two smaller files, and trim duplicate cases in mixed suites. Remove unused snapshots and test scaffolding, then update workflow commands, test selection, documentation, and reliability gates to match. Fix a fish cleanup race, shebang discovery of locally deleted files, and a compiler-output test that counted receipts before the producing compiler finished.

Application behavior and CI job matrices are unchanged. This intentionally retires source-wide grep guards and exhaustive readiness snapshot characterization; related behavior tests do not provide identical static coverage.

Why

Copied inventories, implementation comparisons, source-string assertions, and mock-success smoke tests consume CI capacity while making refactors expensive. Keep independent transcript, real process, recovery, persistence, wire, and ownership tests instead of replacing removed tests with new mirrors.

One matched local runtime cohort improved from 79.2s to 17.6s. Overall PR wall time and queue improvement are not measured yet.

Linked Issue

Related CI work: #25665 and #25760. No new issue.

Visual Proof

N/A — no UI or application behavior change.

Testing

• I manually tested these changes locally
• Automated tests updated; low-value tests deliberately removed

On macOS, before rebase: full mobile 9,964 passed; desktop qualification covers 105,612 passes, one expected failure and 514 existing skips. The final desktop invocation had one unchanged fixture race; its corrected 21-test file passed, and the affected case passed three additional invocations. The failed full-run result remains recorded rather than presented as a single green run.

After rebase onto current main: typecheck, integration/transcript checks and reliability-gate validation run again; normal PR CI validates the integrated branch. Windows/WSL opt-in tests remain, without claiming they ran locally.

Review

Independently reviewed deletions, retained coverage, orphan imports, current workflow references, and preservation of historical native evidence. Dashboard commits are excluded from this PR.

Agent skill upstream boundary

• Not applicable; no upstream skill source or guidance copied.

Notes

Broad deletion diff with one purpose: less low-value test work. SSH, folder workspaces, remote compatibility, and platform behavior stay covered by retained behavioral suites.

Checklist

• This PR is small and focused — focused test cleanup, but a large deletion diff
• Explained the change, mechanism, reason, and tradeoff
• N/A visual proof with reason
• Self-reviewed correctness, security, and performance
• Cross-platform and SSH impact considered
• Complete hosted CI must pass before merge