GitHub experienced multi-service degradation across April 28-29 that knocked GitHub Issues offline for 20 hours 24 minutes, GitHub Pages offline for 20 hours 24 minutes, and GitHub Actions offline for 14 hours 6 minutes, according to GitHub's own status page as tracked by StatusGator. DownDetector reports surged at roughly 9:05 AM CET on April 28. The published incident title was "Incomplete pull request results in repositories." The outage landed in the same week GitHub announced the move to usage-based AI Credits pricing for Copilot — combined, the week makes the "don't depend on a single git host" argument easier to make than it has been in years.
The duration matters more than the trigger. A 20-hour outage on Issues and Pages, and a 14-hour outage on Actions, takes large portions of the developer workflow offline. Issues is where bug tracking and project management live for most open-source projects. Pages is where many small projects host their documentation and demos. Actions is the CI/CD substrate underneath the world's de-facto package distribution graph — when Actions is down, packages don't get published from canonical pipelines, GitHub Container Registry pushes fail, and any project that tied its release process to Actions runners stopped releasing. The single root incident title — "Incomplete pull request results in repositories" — masks the cascade: PR-related metadata feeds into Actions triggers, which feed into Pages deployments, which feed into Issues ticket flows. When the pull-request data layer wobbles, the whole graph above it does.
GitHub is the closest thing to a single-point-of-failure that the developer ecosystem has. There are roughly 200 million developers on GitHub and a non-trivial fraction of the world's open-source CI runs on GitHub Actions. The reliability story this week is uncomfortable: a 20-hour Issues outage is not a 20-minute glitch, and the cause being a PR-data-layer issue suggests the failure was deeper than infrastructure noise. Combined with the Copilot pricing reset to AI Credits announced on April 28, the week reads as "central platform with brittle core systems is also going to start charging more for the AI features." Expect a measurable bump in interest in GitLab, Codeberg, Forgejo, and self-hosted Gitea over the next few weeks. The bump will not be a mass exodus — switching costs are huge — but maintainers of high-traffic open-source projects will start thinking about mirrors and fallbacks more seriously.
For builders, three things to actually do. First, audit your release-process dependency on GitHub Actions specifically: if your package release, container build, or documentation deploy pipeline only fires when Actions runners are healthy, you have a hidden single point of failure. Mirror to a second CI provider (CircleCI, Buildkite, GitLab CI, self-hosted runners) for the critical jobs, even if you keep day-to-day on Actions. Second, push-mirror your repo to a second host. The 90-second setup of a remote on GitLab or Codeberg pays for itself the first time GitHub Issues stops accepting writes. Third, the Copilot pricing reset plus this outage is going to make "AI dev tools that work offline or on alternative infrastructure" a more compelling pitch — watch for product launches in the next month that lead with "your AI coding assistant runs on your infrastructure, not GitHub's."