💡 Merge Queue Readiness: Systematic Prerequisites to Break the O(n²) Re-Review Deadlock

[github-actions[bot]]

Summary

The current merge pipeline forces a full re-review of every open PR on every merge to main, creating an O(n²) review loop that produced 154 forced re-reviews in 48 hours with zero merges (#1864). GitHub merge queue is the proven solution — GitHub's own monorepo processes 2,500+ PRs/month through it with a 33% reduction in wait-to-ship time. This idea systematically addresses the five prerequisites (#1871) blocking adoption: adding merge_group triggers, migrating CodeQL to advanced setup, extending carry-forward verdicts, decoupling review validity from branch-freshness, and retiring the auto-rebase treadmill (#1881).

Market Signal

GitHub reports 33% reduction in wait-to-ship time with merge queue. Industry data (Mergify study, 200K merges across 477 orgs) shows PRs that pass CI still break main 0.77% at 2-5 engineers and 12.5% past 40 engineers — underscoring merge queue value even for small teams. GitHub's May 2026 expansion of concurrency groups to 100 queued runs and third-party merge queue tools (Mergify, Trunk) offering priority lanes and per-label queue rules demonstrate the ecosystem is maturing. The Sep 2026 runner deprecation API also improves queue infrastructure visibility.

User Signal

• Merge protections are unsatisfiable at N>1 open PRs: one merge to main forces a full re-review of every other PR — ~154 forced re-reviews in 48h, zero merges #1864: Merge protections unsatisfiable at N>1 open PRs — 154 forced re-reviews in 48h, zero merges
• Merge-queue prerequisite: no workflow triggers on merge_group, and CodeQL default setup cannot — all five required checks would wedge the queue #1871: Merge-queue prerequisite — no workflow triggers on merge_group, CodeQL default setup cannot
• auto-rebase still merges main into every open PR on every push, though the ruleset no longer requires up-to-date branches — the treadmill's driver outlived the strict fix #1881: Auto-rebase still merges main into every open PR on every push, though the ruleset no longer requires up-to-date branches
• A conflicting PR is never recovered: the rebase sentinel re-fires ~15x per PR reporting success while resolving nothing, and auto-rebase-retry has no manual trigger #1890: Conflicting PRs never recovered — rebase sentinel re-fires ~15× reporting success while resolving nothing
• Auto-Rebase Workflow vs. GitHub Merge Queue #735: Existing Ideas discussion comparing auto-rebase vs merge queue

Technical Opportunity

The engine's carry-forward mechanism (scripts/lib/carry-forward.sh) already handles base-merge verdicts when the diff-vs-merge-base is unchanged (#1870); extending it to queue-rebased commits is incremental. CodeQL can be migrated from default setup to an advanced workflow that triggers on merge_group. The ci-status.sh gating logic already understands required vs non-required checks via ruleset queries. The auto-rebase workflow's merge_group awareness can be feature-flagged to ease the transition.

Assessment

Dimension	Score	Rationale
Feasibility	med	CodeQL advanced-setup migration is the hardest step; other workflow changes are mechanical. GitHub merge queue is GA and well-documented.
Impact	high	Directly resolves the O(n²) re-review deadlock (#1864) and the auto-rebase treadmill (#1881). Estimated savings: 100+ forced re-reviews per week.
Urgency	high	#1864 is actively preventing merges. The auto-rebase treadmill wastes CI budget continuously. Each merge to main compounds the problem for all other open PRs.

Adversarial Review

Strongest objection: CodeQL default setup cannot trigger on merge_group (#1871), and switching to advanced setup requires ongoing manual rule maintenance that GitHub's default setup automates.

Rebuttal: The CodeQL blocker is real but surmountable: advanced setup is GitHub's documented recommendation for organizations needing merge queue. The alternative — the current O(n²) re-review loop — costs more in forced re-reviews and human intervention than CodeQL rule maintenance ever would. Even partial progress (adding merge_group to the other four required checks) delivers value by narrowing the blocker to a single check.

Suggested Next Step

Audit all five required-check workflows (SonarCloud, CodeQL, AgentShield, dependency-audit, duplicate-decl-gate) for merge_group readiness; prototype a CodeQL advanced-setup workflow with the current rule set as a spike to validate feasibility and maintenance burden.

[github-actions[bot]]

Weekly Update

What Changed

Async Merge API GA (Oct 1, 2026): GitHub's async merge API reached general availability, providing a unified, recommended path to merge PRs individually, as stacks, or into a merge queue programmatically. This replaces the older synchronous REST/GraphQL merge calls. (GitHub Changelog, Oct 1 2026)

This is directly relevant to the O(n²) re-review deadlock this idea describes: the async merge API supports programmatic merge queue enrollment, which means the PR-review pipeline could enqueue approved PRs into the merge queue via API rather than waiting for manual merge. Key capabilities:

• Stack merging: the API supports merging PR stacks atomically, which connects to the 💡 Stacked PR Awareness for Review and Dev-Lead Pipelines #792 (Stacked PR Awareness) idea
• Queue integration: PRs can be enrolled into the merge queue programmatically, removing the manual step after approval
• Error handling: the async API returns structured errors for merge conflicts and queue failures, enabling automated recovery for the conflicting-PR recovery gap (A conflicting PR is never recovered: the rebase sentinel re-fires ~15x per PR reporting success while resolving nothing, and auto-rebase-retry has no manual trigger #1890)

Copilot Agent merge (preview, Sept 2026): GitHub's Copilot in VS Code now has "Agent merge" that autonomously handles review feedback, failed checks, merge conflicts, and workflow reruns — demonstrating the platform vendor's commitment to automated merge pipeline completion.

Updated Assessment

Dimension	Previous	Current	Delta
Feasibility	med	high	↑ — The async merge API provides the programmatic merge path; the merge queue itself has been GA since March 2025
Impact	high	high	→ — The O(n²) re-review deadlock remains the most significant throughput bottleneck
Urgency	med	high	↑ — The async merge API is GA now; #1890 (conflicting PR never recovered) and daily stalled-PR reports show the merge path is actively broken

Recommendation

Advance. The async merge API GA removes the last platform prerequisite for merge queue adoption. The suggested next step from the original idea — "systematic prerequisites audit" — should now include: (1) enabling the merge queue on the repo's ruleset, (2) wiring the PR-review pipeline's post-approval step to call the async merge API instead of waiting for manual merge, and (3) validating that the re-review behavior under merge queue is bounded (no O(n²) cascade).