GitHub Actions Down for 6+ Hours: Second-Longest Outage Ever

On August 6, 2026, GitHub Actions — the CI/CD backbone for millions of developers — went down for over 6 hours in what became the platform's second-longest outage on record. And it wasn't just Actions.

The incident started at 15:22 UTC and was still unresolved at 21:30 UTC. The blast radius extended well beyond Actions: GitHub Pages, Copilot code review, Copilot coding agent, and GitHub Enterprise Importer all felt the impact. Webhooks were throttled to ~15% of normal throughput, meaning most pushes and pull requests never triggered any workflow at all.

What Happened

Based on the GitHub Status timeline:

Time (UTC)	Event
15:22	First report: Actions "degraded performance"
15:41	Downgraded to "degraded availability"
15:45	Workflow runs failing to start or failing mid-execution; Actions REST API returning errors; unexpected rate limiting
16:27	Pages begins experiencing issues
17:02	Copilot code review, Copilot coding agent, and GitHub Enterprise Importer affected
17:40	Multiple mitigations applied but not yet resolved
18:46	Jobs still failing or delayed; self-hosted runners hitting errors during registration
20:34	Only ~15% of webhooks processing; ~65% of jobs succeeding (up from 30-40% at the worst point)
21:30	Runners assigned jobs that no longer exist; fix being deployed; GitHub Enterprise Importer paused

A critical detail: this outage affected self-hosted runners too — the very thing many teams invest in to avoid dependency on GitHub's infrastructure. When the Actions control plane has issues, your runners — no matter how well-provisioned — can't pick up jobs.

Not a One-Off

If this were an isolated incident, the story would end here. The statistics tell a different story:

July 2026: 26 incidents
June 2026: 23 incidents
May 2026: 23 incidents
April 2026: 26 incidents
First 6 days of August 2026: 6 incidents

According to Pingoru, GitHub Actions alone has suffered 24 outages since March 2026, totaling 49 hours of downtime. A similar Actions outage occurred just 8 days earlier on July 29, with REST API timeouts and runner registration failures.

Mitchell Hashimoto, creator of Ghostty, publicly stated that GitHub is "no longer stable enough to host serious work" — a damning assessment from one of the most respected developers in open source.

The AI Factor: Is Copilot Eating GitHub's Own Infrastructure?

GitHub has pointed to surging AI usage as a contributing factor, straining shared infrastructure. Copilot's coding agent — which needs significant compute for agentic tasks — competes directly with Actions for the same resources. When both run on the same platform, explosive growth in one can drag down the other.

This isn't a new problem. In May 2026, Microsoft had to call on AWS to help handle AI-related overload. Even with all of Azure behind it, GitHub is struggling to scale for AI workloads.

But blaming AI entirely misses the point. Five consecutive months of 23+ incidents point to a systemic reliability problem, not a temporary one. A platform handling CI/CD for millions of developers can't average ~8 hours of Actions downtime per month without eroding trust.

GitHub apologized in April and promised improvements. In June, SVP Jakub Oleksy committed to changes that would "permanently remove failure modes." The incident numbers since June suggest those changes either haven't been deployed or haven't been enough. The real fix likely isn't patching individual failure modes — it's separating AI and CI/CD infrastructure so growth in one doesn't take down the other.

What This Means for Developers

1. Don't Put All Your Pipelines in One Basket

If your entire CI/CD pipeline depends on GitHub Actions, today's outage could block all deployments. Consider a multi-provider strategy:

GitHub Actions for CI (test, lint, build)
GitLab CI or Gitea Actions as CD fallback
Trigger webhooks to both providers simultaneously

2. Self-Hosted Runners Aren't a Silver Bullet

This incident affected self-hosted runners too — the exact solution many teams adopt to reduce dependency on GitHub. When the Actions control plane fails, even the most powerful self-hosted runner sits idle with nothing to process.

3. Monitor Status Proactively

Subscribe to the GitHub Status RSS feed
Pipe GitHub Status alerts into your team's Slack or Telegram
Set up a fallback trigger: if no webhook arrives from GitHub within X minutes → activate backup pipeline

4. Consider Pull-Based GitOps

Instead of push-based CI/CD (webhooks triggering pipelines), pull-based models like Flux or ArgoCD offer a more resilient alternative. When GitHub webhooks die, a pull-based controller keeps polling and syncing on its own.

Key Takeaways

GitHub Actions went down for 6+ hours — from 15:22 to 21:30 UTC on August 6, 2026, the platform's second-longest outage on record
The blast radius spread wide — GitHub Pages, Copilot code review, Copilot coding agent, and GitHub Enterprise Importer all felt the impact; webhooks dropped to ~15% throughput
Self-hosted runners weren't spared either — when the Actions control plane fails, even self-managed runners can't pick up jobs
Not a one-off — this is the 24th outage since March 2026, totaling 49 hours of downtime
AI is a factor, but not the whole story — Copilot's coding agent competes with Actions for infrastructure, but 5 straight months of frequent incidents points to a systemic problem
Developers should plan for a Plan B — consider a multi-provider strategy, monitor GitHub Status proactively, and look at pull-based GitOps to reduce webhook dependency

GitHub Actions is an excellent tool — when it works. But 49 hours of downtime over 5 months, 26 incidents per month, and a 6+ hour outage that took Pages and Copilot down with it point to a platform struggling with reliability at a structural level.

Developers don't need to abandon GitHub. But it's time to seriously plan for a Plan B. Because the next time you're pushing an urgent hotfix at 2 AM, "GitHub Actions is down" isn't the answer you want to give your team.

Content assisted by AI (Amy 🌸). Reviewed by the author.