On April 23, GitHub had a bad week. A bug in its merge queue feature caused code changes to be silently reversed for more than 2,000 pull requests across 230 repositories.
Today, GitHub's VP of Engineering published a detailed post-mortem acknowledging both this incident and a second separate outage, explaining what went wrong and what the company is doing to fix it.
Here is the version that matters for small business owners who are not developers by trade.
What Actually Broke
GitHub has a feature called a "merge queue." It is designed to help busy development teams safely merge multiple code changes at once without conflicts. It is popular with larger projects that have many contributors pushing updates simultaneously.
On April 23, a bug in this system caused specific types of merges - specifically those using what developers call "squash merge" with more than one pull request in the queue - to produce incorrect results. Instead of adding the new code changes, some merges inadvertently reversed changes from previously merged work.
Translation: imagine asking your developer to add a new feature to your website. They do the work, it gets approved, and then GitHub quietly undeletes the old version and erases the new feature - without any visible error message. That is approximately what happened to the affected repositories.
GitHub said 230 repositories and 2,092 pull requests were affected during the incident window.
Why GitHub Acknowledged a Bigger Problem
The merge queue bug was not the only thing GitHub admitted this week. The company said it is dealing with a capacity crisis driven by what it calls "agentic development workflows" - meaning AI coding tools like GitHub Copilot, Cursor, and OpenAI Codex are generating code at a scale that far outpaces what GitHub's infrastructure was designed for.
GitHub planned to increase capacity by 10X in October 2025. By February 2026, it concluded it actually needs 30X today's scale to handle the load. The company is now migrating systems to public cloud infrastructure and moving performance-sensitive code from its older Ruby codebase to Go.
The short version: GitHub got bigger faster than it planned for, and the cracks are showing.
What This Means If You Hire Developers
If you employ a developer or freelancer who maintains your website, app, or internal tools on GitHub, here are the practical questions worth asking:
Does your developer use merge queue? This specific bug only hit projects using that feature. Most small business codebases do not have enough concurrent contributors to need it, which means you were probably not affected. Worth a quick check.
Do you have any visibility into what actually deployed last week? GitHub's bug produced no obvious error messages. Changes were reversed silently. If something in your app or website stopped working around April 23, this is a possible explanation.
What is your rollback plan? GitHub keeps a full history of all code changes. A competent developer can identify and restore lost work. But if you do not have a developer you can call for exactly this situation, it is worth establishing that relationship before the next outage.
Do you get notified when code changes happen? Tools like GitHub's own status page (githubstatus.com) and basic webhook alerts can tell you when something is wrong before you hear about it from a customer.
The Larger Pattern
GitHub serves more than 150 million developers. It is the default infrastructure for the majority of the world's software projects - including the tools, apps, and websites that small businesses run on.
When GitHub has reliability problems, the effects ripple outward invisibly. A form that stops submitting. A payment integration that breaks. An API that returns errors. Most business owners never know the cause.
The good news from today's post-mortem: GitHub is scaling aggressively, being transparent about the problems, and prioritizing availability over new features. That is the right priority order.
The less good news: the pace of AI-generated code is pushing every infrastructure provider to its limits right now. GitHub is the most visible example, but it is not the only one.
For small businesses, the practical response is not panic - it is making sure you have a developer on call and know where your code lives.
Sources: GitHub Engineering Blog, "An Update on GitHub Availability" (April 28, 2026). Hacker News thread #47932422.