[Delivery System Gap] Consolidate the open PR queue into product verticals and retire superseded micro-PRs

[seonghobae]

Buyer and engineering problem

At the 2026-08-21 08:46 KST live REST snapshot, TEPP had 94 open pull requests: 19 drafts and 75 non-drafts. The split is computed from the same /pulls?state=open&per_page=100 response and is bounded to protected-main SHA 7c29e7c971d7940e1fb3def1ed3aae2d1bc8ad4a; exact-head checks and reviews remain live GitHub evidence. Many PRs are scientifically careful bounded slices, but the queue contains stacked predecessors, replacement vehicles, one-crate-per-rule fragments, stale branch assumptions, and PR descriptions that name a preferred merge that has already changed. A buyer cannot distinguish the small protected-main product from the much larger unmerged design/implementation inventory.

At this snapshot, the exact-head sweep found no unresolved non-outdated review threads. Only PR #146 and PR #148 had failed current-head Checks: both were Strix Caido guest-bootstrap loopback failures with Vulnerabilities 0; the superseded leaf repair #1181 was closed, and the canonical central owner repair is tracked in ContextualWisdomLab/.github#1153 at its live head. Queued Checks remain pending evidence, not a pass.

This queue is itself a release blocker:

• exact-head reviews and checks are spread across too many branches;
• active branches repeatedly become stale or conflicting as protected main advances;
• closely related equations/guards are split into many packages and PRs instead of coherent product boundaries;
• README/TRD maturity statements lag protected-main reality;
• the product-development scheduler can continue producing bounded work without first creating a landable vertical.

Required outcome

Create and execute one authoritative delivery-consolidation plan that reduces the queue without discarding unique scientific evidence.

Acceptance criteria

Inventory and authority

• Enumerate every open PR with exact head/base SHA, draft state, mergeability, current-head checks, review decision/threads, changed paths, dependency/stack parent, owning ADR/requirement, and buyer outcome.
• Classify each PR as landing_vehicle, stacked_dependency, fold_into_landing_vehicle, superseded, duplicate, research_lineage_only, or blocked_external with evidence.
• Record the result in docs/product-technical-gap-baseline.md; the live GitHub API remains authoritative over the snapshot.
• Preserve remote-agent commits and intent; no force-push or closure occurs without first comparing exact remote heads and unique diffs.

Consolidation

• Select a small ordered set of product-vertical landing vehicles aligned to issues [Product Completion] Deliver an executable end-to-end TEPP analysis run #166–[Commercialization Gap] Productionize tenancy, durability, observability, release, and support #174 and existing issue [Interchange Gap] Publish a versioned completed analysis-run result contract #156.
• Fold compatible micro-PRs into the owning bounded crate/service instead of shipping one crate per enum/refusal/equation where that fragmentation creates no reusable boundary.
• Replace stale/conflicted branches with clean current-main replays only when unique implementation/test evidence cannot be merged safely.
• Close superseded/duplicate PRs with a comment naming the replacement and preserved evidence; do not leave ambiguous parallel authorities open.
• Resolve or explicitly disposition every valid review thread before merge/closure.
• Do not treat queued, skipped, cancelled, predecessor-head, or stale checks as passing.

Automation changes

• While the queue exceeds the configured threshold, the hourly product-development workflow prioritizes review repair, conflict resolution, stack consolidation, and merge preparation instead of opening unrelated micro-PRs.
• The scheduler cannot mass-undraft PRs, create empty commits to requeue checks, or change a declared preferred merge without re-reading the live queue.
• Contextual-orchestrator may propose consolidation, but independent verification, publication, review, and merge authority remain separate.
• Add regression tests that prevent PR descriptions/docs from declaring merged PRs active, stale preferred-merge numbers, or active-PR capabilities as protected-main.

Documentation and release truth

• Reconcile README.md, docs/TRD.md, docs/TRACEABILITY.md, ADR maturity, CHANGELOG, and the gap baseline with the exact protected-main implementation.
• The README provides one tested buyer/developer entry path and no longer describes the repository as only an initial empty Task 1 skeleton when domain behavior is present.
• The remaining open queue is small enough that each PR has a clear buyer outcome, dependency, reviewer, and exact-head merge gate.
• The queue reaches zero before GA, or every remaining PR is explicitly release-excluded with a time-bounded rationale.

Queue targets

Use progressive targets rather than closing work arbitrarily:

1. Triage complete: 100% of open PRs classified.
2. Authority unique: no requirement has multiple unexplained landing vehicles.
3. Active queue bounded: at most one ready landing vehicle per product vertical; other dependent work remains a documented stack or is folded/closed.
4. Release queue: zero unresolved review/change requests, zero stale preferred-merge claims, zero superseded duplicates.
5. GA: zero open release-blocking PRs and issues.

Scope boundary

Do not close scientifically unique work merely to reduce a number. Preserve provenance through commits, replacement PR descriptions, issue links, and doctoring. Conversely, local tests and careful prose do not justify retaining dozens of overlapping unmergeable PRs indefinitely.

Related work

• Existing live gap-baseline PR: docs: define TEPP product completion and technical gap baseline #164.
• Hourly contextual-orchestrator workflow: feat: route hourly product agents through contextual orchestrator #165.
• Current result-contract issue/PR: [Interchange Gap] Publish a versioned completed analysis-run result contract #156/feat(api): publish completed analysis-run result contract #157.
• Product-completion issues: [Product Completion] Deliver an executable end-to-end TEPP analysis run #166–[Commercialization Gap] Productionize tenancy, durability, observability, release, and support #174.
• LineageWeave stack: feat(api): serve naruon POSTs on loopback with a live deadline #107 → feat(api): admit LineageWeave as a modular analysis-run consumer #155 → feat(api): expose temporal evidence context for LineageWeave Ask #158/feat(api): project LineageWeave evidence into cutoff-safe histories #159.

Completion evidence

The final PR for this issue must include before/after queue statistics, classification manifest, closed/replaced mappings, protected-main SHA, exact-head check/review evidence for landed PRs, and a refreshed docs/product-technical-gap-baseline.md. No release claim is permitted solely from queue reduction.

[seonghobae]

Live queue and governance evidence refresh (2026-08-21T13:13:05Z UTC / 22:13:05 KST): REST /pulls?state=open&per_page=100 returns 94 open PRs = 19 drafts + 75 non-drafts, against protected-main SHA c45be17. Ruleset 18156473 is active with two required approving reviews, last-push approval, required-thread resolution, and seven main-pinned required workflows; bypass_actors is empty. This explains why local repair commits for PRs #158, #159, and #178 cannot be published by normal branch push, while no protected branch bypass or force push is authorized. Treat this as control-plane evidence: no ordinary feature PR qualifies for guarded force merge, and queued/cancelled checks or absent approvals remain non-passing. The PR/issue consolidation acceptance criteria remain open.

[seonghobae]

Main non-draft PR queue reached zero at 2026-08-25T02:45Z.

Consolidation completed through three mechanisms:

1. Hourly scheduler independent merges (~40+ slices)
2. Batch integration vehicle chore(delivery): consolidate 31 queued green slices into one landing vehicle (#175) #215 (31 folded green slices with per-slice provenance)
3. Individual rebase-admin-merge passes for stragglers

The 31 remaining open PRs are all drafts forming stacked psychometric recovery chains that follow the stacking plan — not independent landing vehicles.

GAP-012 is closed.