Downloads Failed at Twice the Rate of Everything Else
The Microsoft-owned developer platform began reporting issues around 13:40 UTC (9:40 AM ET), immediately affecting core services. Pull Requests, Issues, Webhooks, Actions, Copilot, and GitHub Pages all experienced degradation. Web and API traffic saw roughly 20% error rates, but archive and raw repository downloads faced a much steeper 50% failure rate.
| Service | Error Rate / Status |
|---|---|
| Web experiences | ~20% |
| API traffic | ~20% |
| Archive downloads | ~50% |
| Raw repository content | ~50% |
| Copilot | Degraded availability |
| Pull Requests, Issues, Webhooks, Actions | Degraded |
| Git Operations, Packages, Codespaces | Operational |
Developers could not push code, review changes, or run automated workflows during the disruption. Git Operations, Packages, and Codespaces notably stayed operational throughout. User complaints on Downdetector spiked around 13:30 UTC, and frustration mounted quickly across social media. One user put it bluntly: Github is really falling apart. Can we stop being down please. We got work to do.
Three Outages in Three Months
This incident adds to a troubling pattern building since spring. GitHub recorded eight service-degrading events in July alone. May saw nine incidents, and June recorded six. Just a week before this latest outage, GitHub Actions experienced a particularly lengthy disruption where workflow runs failed or sat queued for hours. GitHub later called that specific outage unacceptable.
The company is currently undertaking a major infrastructure migration, shifting operations toward Microsoft’s Azure cloud platform. The stated goal is reducing shared failure points and improving isolation between individual components. Despite that effort already underway, Monday’s failure still managed to hit multiple critical services simultaneously.
One Platform, Many Points of Failure
The outage exposes how deeply GitHub sits inside modern development pipelines. Consider what depends on the platform functioning correctly:
- GitHub Actions powers automated testing and deployment
- Webhooks connect the platform to countless external services
- Pull Requests underpin the entire code review process
- GitHub’s APIs are woven into third-party development tools
A failure in any of these areas can halt an organization’s software delivery pipeline entirely, even when basic Git operations keep working underneath.
AI Coding Is Adding New Load
Growing scrutiny of GitHub’s reliability record points to increased infrastructure strain from a specific source: AI assisted coding and agentic workflows. Microsoft CEO Satya Nadella claimed last year that AI writes up to 30% of code in some of the company’s repositories, subject to human review. Addressing repeated outages back in June, software engineering SVP Jakub Oleksy said structural changes were already underway, pledging to permanently remove failure modes.
Hashlytics Take
The pattern here is worth sitting with. GitHub is mid migration to Azure specifically to reduce shared failure points, and it still had a day where five separate services degraded at once. That is not evidence the migration is failing. It is evidence that the underlying load has grown faster than the fixes can keep pace, and AI assisted coding at 30% of some repositories is a meaningful part of that load. Every organization treating GitHub as invisible infrastructure should be asking what their actual exposure looks like on a bad day, because “bad days” are now running at roughly one every four days this summer.
Follow Hashlytics on Bluesky, Facebook, LinkedIn , Telegram and X to Get Instant Updates
