GitHub’s Actions service hit a snag on May 5, 2026. Elevated queue times and failures plagued jobs on hosted runners in the East US region. The incident, tracked at GitHub Status, started at 13:37 UTC with reports of degraded availability. By 13:48 UTC, engineers pinpointed the issue: elevated queues on standard hosted runners, impacting 10% of runs.
Updates rolled in steadily. At 14:14 UTC, failures joined the queue delays, affecting 8% of runs. Private networking runners could failover to other Azure regions for relief. Two hours later, at 15:12 UTC, the team was coordinating with their compute provider—widely known to be Azure—to ease the strain on 10% of affected jobs. Mitigation landed by 15:54 UTC for standard runners. Monitoring kicked in. Private networking setups in East US lingered under capacity shortfalls.
No root cause spilled publicly yet. But patterns emerge. GitHub depends on Azure for these runners. Regional capacity limits there often cascade. Developers saw jobs stall. CI/CD pipelines froze. Builds backed up.
This isn’t isolated. GitHub’s uptime has eroded. A reconstructed status page shows dips below 90% in 2025, April 2026 under 85%. HashiCorp co-founder Mitchell Hashimoto vented frustration last week, calling GitHub “no longer a place for serious work” after near-daily outages blocked PR reviews for hours. “Almost every day has an ‘X’,” he journaled, per The Register.
Outages Pile Up as AI Demand Surges
April brought catastrophe. A merge queue bug reverted commits in 2,092 pull requests—a data integrity nightmare. GitHub downplayed it as hitting 0.07% of customers. Developers fumed. Gergely Orosz, ex-Uber engineer, slammed the response on X: “Massive L from GitHub… No respect for the customer.” Tom Warren of The Verge noted the timing: hours after his report on internal reliability gripes, the platform crumbled.
Actions has faltered before. February saw delays in notifications, Copilot policy propagation. April 28: Elasticsearch snafus stalled PRs. Copilot itself strained under AI boom. GitHub imposed rate limits, suspended new sign-ups, shifted to metered billing by June 1. “High concurrency and intense usage” taxed infrastructure, they admitted.
Why now? AI tools exploded demand. Copilot generates code cheap. But verification lags. One X user tallied 257 outages since some baseline, 83 from capacity. Projects flee. Zig migrated to Codeberg over AI nags and decline. Ghostty plans exit. Gentoo ditched mirrors for Copilot pushes. Even OpenAI eyes a rival repo, spurred by outages, says The Verge.
NHS acts today. Internal guidance demands privatizing public repos by May 11 over AI scraping and security fears from Anthropic’s takedowns, per fresh Register reporting. Thousands of repos vanished in April when Anthropic’s DMCA overreach hit forks, TechCrunch detailed.
GitHub apologized broadly last week for slipping uptime. “We are sorry,” they posted, vowing fixes amid three-nines struggles, as The Register covered. A recent RCE vuln got patched in hours via AI-hunted flaw, The Verge reported—no exploitation.
Failures sting in dev workflows. Pipelines halt. Releases delay. Teams scramble to self-hosted runners or rivals like GitLab. East US bias hurts US-East devs hardest. Failover helps some. Others wait.
GitHub scales frantically—30x expansion mid-flight, one X post claims. Azure ties bind tight. Capacity chokes persist. Devs demand better. Hashimoto’s journal resonates. Outages compound. Trust frays.
Mitigation holds for now. Full recovery? Eyes on status page. Broader fixes? Microsoft owes details. Actions powers millions of workflows. One region falters. Ripples spread.
WebProNews is an iEntry Publication