Service eligibility and regional restrictions
PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.
Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.
Key Takeaways
- The incident began October 5, 2026 at 19:11:58 UTC—October 6 at 03:11:58 Beijing time. A19:15:17 UTC update described hosted-runner assignment delays.
- At 20:39:27 UTC GitHub included job failures and delays in its investigation. At 20:47:22 UTC Actions became major_outage and impact critical; those are historical states.
- Mitigation was applied at 21:32:31 UTC. Actions was normal at 21:54:23, Pages at 22:40:41, and the incident formally resolved at 22:49:42. At 23:24:31 UTC it was resolved with both components operational.
- Check the actual original run, jobs, and steps. queued means neither failed nor passed tests.
- Keep the original run link, commit, and times; avoid unsupported redispatch. Retain existing tests, reviews, and release gates.
The incident is resolved; your original run still needs checking
For queued Actions runs that never started, the latest official update resolved this incident at 22:49:42.660 UTC on October 5, 2026—06:49:42 Beijing time on October 6. Actions and Pages had recovered separately beforehand; still check your workflow’s actual state. “Incident with Actions” began at 19:11:58.373 UTC, or03:11:58 Beijing time October 6, investigating reports of degraded Actions performance.
The19:15:17.486 UTC update described delays assigning GitHub-hosted runners to Actions jobs. Workflows across runner configurations could take longer to start. The team was mitigating and would update as it learned more. Mitigation in progress did not mean completed mitigation or recovery.
At 19:50:50.720 UTC GitHub continued investigating hosted-runner assignment delays affecting startup across runner configurations and continued mitigation.
At 20:39:27.955 UTC it expanded the investigation to job failures and delays affecting hosted-runner assignment and startup. An earlier20:46:12 UTC read showed minor impact and Actions degraded_performance; this is a historical snapshot.
At 20:47:22.557 UTC GitHub again reported degraded Actions availability and ongoing investigation, changing the component to major_outage and impact to critical. This historical escalation is not the current component state; labels are not measured failure rates.
At 21:09:15.654 UTC, alongside Actions/Hosted Runners issues, some customers could not access repository lists, licensing, and billing pages; investigation and mitigation continued. At 21:22:39.017 UTC Pages degradation was added. This did not claim all GitHub pages were inaccessible.
At 21:31:18.177 UTC investigation continued; Actions moved from major_outage to degraded_performance and Pages was degraded_performance. At 21:31:43 UTC the incident remained investigating/critical, with monitoring and resolved times empty. This snapshot predates the next mitigation update.
At 21:32:31.460 UTC GitHub said mitigation for Actions failures was applied: queued jobs were clearing, new jobs no longer delayed, and full runner recovery was expected soon. It continued watching recovery and mitigating repository-list, licensing, and billing access issues. At 21:36:19 UTC the incident was still investigating/critical and Actions/Pages degraded_performance. Monitoring in prose did not mean the API had switched to monitoring.
At 21:54:23.028 UTC Actions was explicitly normal and changed to operational. At 22:08:39 UTC Pages remained degraded_performance and the incident investigating/critical, with monitoring/resolved times empty.
At 22:40:41.795 UTC Pages was explicitly normal and became operational while Actions stayed operational. At 22:44–22:45 UTC the incident still read investigating/critical with empty monitoring/resolved times. The summary showed All Systems Operational before the incident formally closed; these states must remain distinct.
At 22:49:42.660 UTC the official update formally resolved the incident and promised detailed root-cause analysis when available. This article read the incident API, original page, and summary at 23:24 UTC: resolved_at matched that update, Actions/Pages were operational, and All Systems Operational had no active incidents. The retained critical impact field does not mean a continuing severe outage. Resolution is not individual verification of every account page or run.
What was confirmed, and what remains unknown
Earlier scope included job failures/delays affecting hosted-runner allocation and startup, Pages degradation, and some customers’ repository-list, licensing, and billing access. The latest text resolves the incident. No complete list of runner operating systems, types, regions, repository categories, or plans was published. Resolution does not mean every older run finished or every account page was checked.
Full runner recovery soon was a21:32 expectation; Actions normal at 21:54, Pages normal at 22:40, and resolution at 22:49 were subsequent confirmed updates. The incident page read lacked the promised detailed root-cause analysis, waiting-time compensation, or results for every backlog run. Causes from other incidents cannot explain this one. A contemporaneous queue is only an observation until run/job evidence establishes fit.
Separate queued, running, and test conclusions
Treat queued as a waiting state to investigate, not tests failed. An unstarted job has not run its tests; a longer wait is not a test result, and waiting does not prove code passed.
Distinguish the whole workflow run from each job. Other jobs may have their own states and conclusions while one queues. Read the run summary, jobs, and steps. One waiting job does not describe the entire run, and a final failure is not automatically caused by runner allocation.
The shortest read-only check: open the original run
GitHub’s history guide requires repository read access. Open the repository, select Actions, choose the workflow on the left, then open the original run’s summary. Confirm commit, branch, and trigger time so you do not inspect an earlier run.
Logs contain job and step status. Compare the summary and saved logs to record waiting, executed, and concluded jobs. If a job never started, absent test output means not executed; do not invent a failure cause or passing result.
Save run IDs and times instead of duplicating queued work
Save the original link/run ID, target commit, workflow/job names, runner configuration, first queue observation, and subsequent check times. Mark UTC or local timezone to compare with the19:11:58 UTC start. Keep redacted summaries without credentials or private logs.
Establish whether the original run waits, has started, or has a terminal result before acting. Triggering the same workflow without evidence makes authoritative results ambiguous; deployment or other external actions can also be duplicated.
Keep CI and release gates while waiting
If runner waiting delays release, record waiting for execution. Unexecuted tests have not passed. Retain tests, human review, required checks, and release approval; server delay does not justify weakening them.
After official updates, check the same run again. Monitoring or recovery does not prove every backlog run finished and cannot replace tests. If actual execution yields failure logs, investigate those specific failures rather than attributing all of them to this incident.
What to watch next
Watch for detailed official root-cause analysis and whether original jobs execute and conclude. Record service incident state separately from your run: one explains how the incident ended, the other how far this workflow actually ran.
At this check the incident is formally resolved and Actions/Pages normal; individual run results and account pages still need checks. This is incident reporting and a minimal read-only sequence, not testing of readers’ repositories or proof of a queued run’s cause. Follow official analysis and your actual results.
Sources
Frequently Asked Questions
Does an Actions run stuck queued mean tests failed?
No. Check the full run, job, and steps. An unstarted job has not run its tests, and other jobs need separate checks.
Has this incident recovered?
GitHub resolved it October 5, 2026 at 22:49:42.660 UTC—October 6 at 06:49:42 Beijing time. At 23:24:31 UTC it was resolved, Actions/Pages operational, and no active incidents in the summary. Detailed root-cause analysis is forthcoming. This does not replace account-page or original-run testing checks.
Were all runners confirmed affected?
No configuration list or proportion was published. Earlier scope included hosted-runner allocation/startup failures and delays, Pages, and some account pages. Resolution does not establish completion of every old run or verification of all account pages.
Should I redispatch a queued workflow?
First preserve and inspect the original run; waiting alone is not a reason to duplicate it. For deployment/external actions, establish original outcomes and execution rules first.
Can I skip CI after the status page recovers?
No. Service recovery does not replace your test results. Existing checks, reviews, and release gates must pass before following the normal release process.
Does a queue at the same time prove the same cause?
No. Timing is only an observation; check run, job, runner configuration, and actual state. The read page lacked promised root-cause analysis, full scope, and each backlog result. Do not invent a diagnosis for a personal run.