ORA

Pattern

ORA-2026-0065

ID
ORA-2026-0065
Date
2026-04-19
Status
unknown
Maturity
M2
Source
docs/entries/patterns/ORA-2026-0065_dispatch-language-drifts-from-deploy-topology.md

fleet-ops

Pattern

Fleet dispatch posts describe Heartwood production landings as "merge to main + deploy" while the ratified deploy topology (FLT-0014 Option B, codified in heartwood/docs/ops/heartwood-deploy-topology.md) says production deploys from codex/hwd-0015-ready. Implementers who read the dispatch literally would create a false-main merge that doesn't reach production.

Evidence

  • FLT-0175 (2026-04-18): dispatch text said "Merge FLT-0083 worktree -> main + deploy"
  • FLT-0179 (2026-04-18): CAMBER-05 caught the mismatch mid-execution, honored the real topology instead of the dispatch text, and filed ORA-0030

Root cause

Dispatch authors (both human-prompted and shepherd-generated) compose merge+deploy instructions from mental models formed before FLT-0014 ratified the feature-branch deploy model on 2026-04-16. The topology decision lives in a docs/ops file that isn't loaded into dispatch composition context.

Why this matters

A seat that follows dispatch language literally would: 1. Merge to main (stale, not deployed by Vercel) 2. Possibly run vercel --prod from main (topology violation) 3. Or report "deployed" without the code actually reaching production

CAMBER-05's self-correction cost ~5 min of in-session investigation. A less-aware seat would have created a silent non-deploy.

Audit result (2026-04-19)

All active dispatch surfaces are clean:

  • HEARTWOOD_WORK_BACKLOG.md line 43-44: "Production deploys from codex/hwd-0015-ready — NOT main" (correct)
  • Shepherd prompt/protocol: no Heartwood + main references
  • Fleet scripts (lane-boot, feed-append, shepherd-fire): no Heartwood + main references
  • ORA doctrines: no Heartwood + main references
  • heartwood-deploy-topology.md: authoritative, correct
  • FLEET_FEED.md: historical entries exist but feed is append-only; CAMBER-05 already self-corrected

Guardrail recommendation

The current guardrail is the backlog language + topology doc. No active template references main for Heartwood, so the remaining risk is ad-hoc dispatch posts composed without consulting the topology. This is a training/awareness issue, not a template bug. The pattern is self-healing when implementers check the topology doc before executing merge+deploy dispatches, which CAMBER-05 demonstrated.

No additional tooling change recommended at this time. If the pattern recurs, consider adding a shepherd pre-dispatch check that warns when a Heartwood dispatch mentions main without codex/hwd-0015-ready.