Overview
Merge queue that guarantees branch stability and accelerates development at enterprise scale
Trunk Merge Queue is a hosted service that manages and controls how pull requests are merged into your main branch. It ensures incompatible changes never break the branch while maximizing merge throughput—essential for high-velocity teams working in monorepos.
The Problem: Traditional Merge Workflows Break at Scale
Traditional workflows fail at high velocity because of the fundamental race condition between testing and merging:
"Merge when green" breaks
main: PRs pass CI on their own branches but never test integration. At high velocity, this guarantees eventualmainbreakage when independently-passing PRs conflict."Require up-to-date branch" prevents merges: GitHub's safeguard (rebase + retest) creates a race condition. While a PR's CI runs after rebasing, other PRs merge, making it stale again. At high velocity, PRs get stuck in endless rebase-retest loops and can't land.
The result: At high velocity, teams can't deploy reliably. Either main breaks regularly—forcing engineers to investigate failures and coordinate reverts—or PRs get stuck in endless rebase-retest loops, starving the pipeline.
The solution is predictive testing: validating that PRs will pass when combined with the future state of main. This eliminates the race condition—ensuring both branch stability and that PRs can actually land.
How Trunk Merge Queue Solves This
Predictive testing eliminates the race condition
Trunk Merge Queue tests each PR against the predicted future state of main—the state after all PRs ahead of it in the queue have merged. This eliminates the race condition: by the time a PR merges, it has already been validated against the exact state it will merge into.
When PR A and PR B both enter the queue, Trunk tests B against main + A, guaranteeing that the merged result will be stable. No stale test results, no post-merge failures, no broken main.
But predictive testing at scale requires optimization
Testing every PR serially against the predicted state guarantees correctness but creates new bottlenecks: queue wait times and CI costs scale linearly with PR volume. Trunk solves this with enterprise-grade optimizations:
Intelligent Batching runs one CI test for multiple PRs together instead of testing each separately, reducing costs up to 90% with automated bisection when batches fail
Dynamic Parallel Queues analyze which code each PR touches and create independent test lanes for non-overlapping changes—eliminating queue bottlenecks from unrelated PRs in large monorepos
Anti-Flake Protection uses results from later PRs in the queue to validate earlier ones—if a PR fails but subsequent PRs that include its code pass, the failure was likely transient and all PRs merge together
The result: guaranteed branch stability with throughput that scales to thousands of PRs per day.
Performance Guarantee
We stress test our merge queue to consistently merge 250+ PRs/hour over 12-hour periods—supporting 3,000+ PRs per day. This is our validated performance floor, not a ceiling. If your team requires higher throughput, we’ll work with you to validate support for your specific volume. Contact us.
Next steps
Last updated