[SafetyMP]

The comparison previously ended in a worksheet whose recommendations did not guide the investigation. This change connects all five routes around Compare → Change/Inspect → Form a position, with recommendations that match the retained case and outcome.

• Introduce the two control bundles, retain the 61/59 timing comparison, and focus the next meaningful step.
• Show concrete causal stages for all nine cases and distinguish a violated constraint from a held boundary or unresolved review.
• Make recommendations expose relevant evidence requests, review questions, or an eligible hypothetical recovery; attach modeled checks and recorded events without replacing authored notes.
• Explain browser/source relationships before the full hazard register, controls and recorded traces; preserve saved sessions, immutable comparisons, fixtures and authority boundaries.
• Pin the new browser runner and exact 233-check postdeployment result; allowlist only this candidate branch while preserving existing verification protections.

Validation: 303 frontend tests, 233 built-site Chromium checks, 7 simulator tests, 25 production-integrity mutation tests, 136 evidence-verifier tests, lint/typechecks/build/static/fixture/readiness checks passed locally. Independent semantic, design, visual and infrastructure reviews passed.

Linux CI run 43644378045 passed. The strict predeployment gate accepted the original source-bound artifact, exact head cc6aaa8, tested merge 4510ed9, and seven independently inspected Linux screenshots. The Linux browser result also passed all 233 checks with no console errors or nonstatic requests.

These checks establish implementation behavior, not measured user learning. No merge or deployment is included. Implementation and acceptance record: docs/APS_GUIDED_INVESTIGATION.md.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota. Try requesting a review at a lower effort level.