BAD CODE: separate fast PR iteration gates from release-grade verification

[flyingrobots]

Problem

Normal PR iteration is now paying release-grade verification multiple times:

• local pre-push runs Code Dojo debt/strict checks, DOGFOOD i18n completeness/build/debt, code size, typecheck, build, full chunked Vitest, and scripted interactive example smoke;
• GitHub then runs overlapping remote gates: Code Dojo, full tests on Node 20 and Node 22, focused unit tests on Ubuntu and Windows, DOGFOOD smokes, Bench Gradient, changed-path classification, and CodeRabbit;
• a small standards-only slice on PR Respecting the Dojo: 1000-count burndown #375 still had to wait for full local verification plus the full remote matrix.

This is not a request to weaken release quality. It is workflow debt: the repo lacks a clear fast PR iteration lane versus a full merge/release-grade gate.

Evidence

Observed during PR #375 at commit 6bb5b8b0:

• local npm run code-dojo:ci passed only after build, workspace lint, typecheck, ESLint, and eight Vitest chunks;
• git push repeated a heavy pre-push gate including full tests and scripted interactive smokes;
• GitHub checks then reran remote matrix jobs, with Node 20/22 full tests and Code Dojo taking the longest visible wall time;
• all checks were green, but the iteration loop spent most of its time proving broad gates that were not specific to the docs-preview helper slice.

Related narrower debt:

• BAD CODE: split smoke gates from real PTY and subprocess dependencies #299 tracks smoke gates depending on real PTY/subprocess behavior.
• BAD CODE: replace repeated DOGFOOD scripted renders with fixture-backed view harness #298 tracks repeated DOGFOOD scripted renders.
• BAD CODE: Code Dojo debt reporting hides live slack under stored ceilings #369 tracks Code Dojo reporting visibility.

Desired Shape

Design a two-lane verification model:

1. Fast PR iteration lane keyed from changed paths and risk class.
2. Full merge/release lane that preserves current release-grade guarantees.
3. A no-surprises local workflow where pre-push does not duplicate every expensive remote job by default.
4. Clear escalation rules for touching core runtime, generated assets, DOGFOOD sources, release docs, or ratchet baselines.
5. Timing evidence in CI summaries so slow gates are visible instead of anecdotal.

Acceptance Criteria

• Document the intended fast/full gate split before implementation.
• Keep Code Dojo ratchets enforceable.
• Keep release-grade full verification available and required where appropriate.
• Make a typical docs/test-only ratchet slice substantially faster to push and review.
• Add timing output or a CI summary that shows which jobs consumed the most wall time.

[coderabbitai]

🔗 Related PRs

#43 - feat: share release validation tooling [merged]
#44 - feat: add merge readiness tooling [merged]
#320 - Strengthen Bijou release policy [merged]
#330 - Start DX-046 DOGFOOD GraphQL fixture [merged]
#337 - Polish DOGFOOD shell and Blocks defaults [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!

[flyingrobots]

Backlog release triage: assigned to v8.2.0.

Reason: Fast PR gates versus release-grade verification is Method and CI hardening after the v8 contract/replay lanes settle.

This keeps the GitHub issue tracker aligned with the expanded post-v8 release horizon instead of leaving active work in Beyond or without a milestone.