Resolves #13978 Related: #13755

Golden Path now treats semantic-route failures as failed task outcomes instead of successful empty runs. Chroma query failures render a fresh degraded Computed Golden Path section, prune stale frontier GUIDES, and return a structured failure outcome; the scheduler records that outcome via markFailed and health telemetry instead of refreshing lastSuccessAt.

Evidence: L2 focused unit coverage -> L2 required for the Golden Path scheduler and handoff-rendering close target. Residual: none for #13978.

Deltas from ticket

The fix also covers local graph-store mapping failures with the same structured failure path, because those would otherwise create the same false-green class after semantic candidates are found.

Test Evidence

• npm run test-unit -- test/playwright/unit/ai/services/graph/GoldenPathSynthesizer.spec.mjs -> 34 passed
• npm run test-unit -- test/playwright/unit/ai/daemons/orchestrator/scheduling/pipeline.spec.mjs -> 16 passed
• npm run agent-preflight -- ai/services/graph/GoldenPathSynthesizer.mjs ai/daemons/orchestrator/scheduling/pipeline.mjs test/playwright/unit/ai/services/graph/GoldenPathSynthesizer.spec.mjs test/playwright/unit/ai/daemons/orchestrator/scheduling/pipeline.spec.mjs -> passed
• git diff --check -> passed
• pre-commit hooks -> passed

Post-Merge Validation

• Restart the orchestrator and verify a Golden Path route failure records golden-path.lastErrorAt / failed health outcome instead of lastSuccessAt.
• Confirm resources/content/sandman_handoff.md shows a degraded Computed Golden Path section rather than stale numbered recommendations after a route failure.

Authored by Euclid (GPT-5, Codex Desktop). Session unavailable in Codex Desktop harness.