LearnNewsExamplesServices
Frontmatter
titlefix(ai): prevent Golden Path sync starvation (#13750)
authorneo-gpt
stateClosed
createdAtJun 21, 2026, 12:52 PM
updatedAtJul 27, 2026, 12:08 AM
closedAtJun 21, 2026, 12:57 PM
mergedAt
branchesdevcodex/13750-golden-path-bisect
urlhttps://github.com/neomjs/neo/pull/13756
contentTrust
projected
quarantined0
signals[]
Closed
neo-gpt
neo-gpt commented on Jun 21, 2026, 12:52 PM

Resolves #13750

Fixes the Golden Path freeze at the scheduler / lease root instead of adding a filesystem fallback. The shared heavy-maintenance lease now supports explicit compatible-owner execution, so githubWorkflowSync can run while local kbSync owns the lease and refresh issue graph state instead of self-deferring. Golden Path now treats githubWorkflowSync as a graph-producing dependency, so synthesis waits for the sync/ingest path rather than reading stale graph state mid-refresh.

Evidence: L2 focused unit coverage -> L4 post-merge daemon validation required for live cadence. Residual: #13624 still owns the broader net-progress / idle-condition accounting beyond this starvation mechanism.

Deltas from ticket

  • Delivers the heavy-maintenance lease starvation path named in the #13750 bisect.
  • Keeps Golden Path graph-authoritative: no synced-markdown fallback, no handoff rendering workaround.
  • Extends the existing memory-summary-backfill / kbSync compatibility through the child CLI layer too, because the same self-deferral bug class existed there.

Test Evidence

  • node --check ai/daemons/orchestrator/services/HeavyMaintenanceLeaseService.mjs
  • node --check ai/daemons/orchestrator/services/MaintenanceBackpressureService.mjs
  • node --check ai/scripts/maintenance/syncGithubWorkflow.mjs
  • node --check ai/scripts/lifecycle/backfill-memory-summaries.mjs
  • npm run test-unit -- test/playwright/unit/ai/daemons/orchestrator/services/HeavyMaintenanceLeaseService.spec.mjs test/playwright/unit/ai/daemons/orchestrator/services/MaintenanceBackpressureService.spec.mjs test/playwright/unit/ai/daemons/orchestrator/scheduling/registry.spec.mjs test/playwright/unit/ai/daemons/orchestrator/scheduling/picker.spec.mjs test/playwright/unit/ai/daemons/orchestrator/scheduling/pipeline.spec.mjs test/playwright/unit/ai/scripts/maintenance/syncGithubWorkflow.spec.mjs test/playwright/unit/ai/daemons/orchestrator/daemon.spec.mjs — 111 passed after rebase onto origin/dev
  • git diff --check

Post-Merge Validation

  • Restart the live orchestrator and verify githubWorkflowSync runs when kbSync is active or holds the heavy-maintenance lease.
  • Verify issue-to-graph ingestion catches current #137xx issues and Golden Path no longer recommends stale non-board work.
  • Verify the next resources/content/sandman_handoff.md refresh is produced by the graph path, not a fallback.

Commits

  • 6f564fdd9cfix(ai): prevent Golden Path sync starvation (#13750)

Authored by Euclid (GPT-5, Codex Desktop). Session 019ee5c2-82ba-7b73-8812-df59106ff61a.

neo-gpt
neo-gpt commented on Jun 21, 2026, 12:57 PM

Withdrawing this PR before review routing.

Operator correction is valid: this over-claims #13750 by treating scheduler/lease compatibility as the root Golden Path repair. The actual drill target is graph ingestion / Tri-Vector analysis draining the system and preventing a real Golden Path refresh after current top recommendations are resolved. This PR must not be reviewed or merged as a #13750 fix.

No reviewer was requested. I am closing this and moving the lane back to live ingestion / Tri-Vector forensics under #13750.