Incident with Actions, Pull Requests and Issues

Frequent outages affecting GitHub Actions, pull requests, and issues are prompting developers to question GitHub’s reliability and Microsoft’s strategy as AI-driven “agentic” coding massively increases load on the platform. Many argue that generous free tiers and underpriced CI minutes have effectively let bots DDoS the service, and suggest stricter rate limits, higher prices, or separating free and paid usage to protect serious workflows. Others are responding by exploring alternatives like GitLab, Buildkite, Forgejo, or fully self-hosted CI, weighing the trade-off between SaaS convenience and operational control.

Overall sentiment

  • Many commenters are frustrated and no longer view GitHub Actions as reliable for “serious” or enterprise CI/CD.
  • Some still find it acceptable or great for hobby/small/FOSS projects where free, low-effort CI is highly valued.
  • Several teams are actively planning or executing migrations away from GitHub CI.

Perceived causes of repeated outages

  • Massive growth attributed to AI/agentic usage: claims of going from ~50M to ~2B events per day, with agents repeatedly triggering CI for trivial changes.
  • Disagreement over infrastructure:
    • Some blame Azure as “unreliable” or capacity-constrained.
    • Others argue GitHub’s own architecture and scaling decisions are at fault, not the underlying cloud vs. self‑host.
  • Debate on whether outages are due to:
    • Genuine capacity limits and write-heavy workloads (actions, PRs, events).
    • “Vibecoded” / agent-written slop and operational shortcuts.
    • Management choosing growth and AI positioning over stability.
  • Unclear how much of the impact is shared between free tiers vs. paying/enterprise; reports conflict on whether enterprise cloud avoids or shares the pain.

Pricing, free tier, and incentives

  • Many see this as a pricing and policy failure:
    • Free/private repos plus generous free Actions minutes allegedly invited abusive or heavy AI-agent usage.
    • Suggestions include making Actions a paid feature, sharply reducing free minutes, stricter rate limits, or “degraded” service for free users.
  • Counterpoints:
    • Free CI is crucial for open source and user acquisition.
    • Over-aggressive changes could alienate core communities or create “mystery charges.”
    • Low pricing may have been strategic to gain data and network effects.

Alternatives and coping strategies

  • Alternatives mentioned: GitLab CI, Buildkite, Semaphore, Concourse, Forgejo + self-hosted runners, RWX, Azure DevOps, home servers.
  • Arguments for self-hosting: independence, reliability, performance, and reduced platform risk, now made easier with LLM assistance.
  • Arguments against self-hosting: operational overhead, cost, difficulty of Windows/macOS runners, and better use of developer time on SaaS.
  • Some teams throttle CI (manual triggers, only on main, local testing) to reduce load and dependence on Actions.

Broader concerns

  • Worries about a potential “death spiral”: paying users leaving → less investment → more instability.
  • Criticism of the Microsoft acquisition and broader antitrust environment.
  • A few suggest Git itself or GitHub’s architecture may be structurally hard to scale at current event volumes.