GitHub shipped native stacked pull requests on 2026-07-30, and they are merge-queue aware. Independently confirmed at docs.github.com/en/pull-requests/get-started/about-stacked-prs, verbatim:
"Merge the entire stack at once by merging the top pull request. Every pull request below it comes with it."
"Stacks support merge commit, squash, and rebase merge methods, and they are merge-queue aware."
This makes README:23 and PRD:55 materially false as written, and kills END-STATE:64's Bar B anchor "3-PR stack, time-to-land p95 <= 1.3 CI cycles (GitHub: ~3)".
Three of five rows of the README competitive table are false against primary sources — one of them contradicted by our own research-merge-queue-landscape.md, committed the same day:
The surviving differentiators are exactly two, and both are narrower than claimed:
- Validation count — one CI round for the whole stack vs N
- Suffix atomicity — all-or-nothing vs GitHub's per-layer validation, which permits partial landing
Also add the strongest single piece of evidence for the thesis, which appears in none of our docs: GitHub's own 2026-04-23 merge-queue incident — 230 repos, 2,092 PRs silently reverted, 3h33m to detect, found by support tickets. First-party, recent, and exactly the failure mode invariant 1 exists to prevent.
And resolve the delivery-shape contradiction: README:101 says "No customer-operated infrastructure to start... The hosted control plane is ours"; README:134 says "self-hosted OSS control plane first"; PRD:98 says hosted-first; END-STATE:23 says self-hosted-first. Three documents, two mutually exclusive v1.0s.
See docs/INVESTIGATION.md §7 R12, §6 Q12.
GitHub shipped native stacked pull requests on 2026-07-30, and they are merge-queue aware. Independently confirmed at
docs.github.com/en/pull-requests/get-started/about-stacked-prs, verbatim:This makes README:23 and PRD:55 materially false as written, and kills END-STATE:64's Bar B anchor "3-PR stack, time-to-land p95 <= 1.3 CI cycles (GitHub: ~3)".
Three of five rows of the README competitive table are false against primary sources — one of them contradicted by our own
research-merge-queue-landscape.md, committed the same day:max_parallel_checks, cumulative batch PRs) and does bisect (binary search on batch failure)The surviving differentiators are exactly two, and both are narrower than claimed:
Also add the strongest single piece of evidence for the thesis, which appears in none of our docs: GitHub's own 2026-04-23 merge-queue incident — 230 repos, 2,092 PRs silently reverted, 3h33m to detect, found by support tickets. First-party, recent, and exactly the failure mode invariant 1 exists to prevent.
And resolve the delivery-shape contradiction: README:101 says "No customer-operated infrastructure to start... The hosted control plane is ours"; README:134 says "self-hosted OSS control plane first"; PRD:98 says hosted-first; END-STATE:23 says self-hosted-first. Three documents, two mutually exclusive v1.0s.
See
docs/INVESTIGATION.md§7 R12, §6 Q12.