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 September 5 at 04: 39 Beijing time and reached major impact. GitHub entered monitoring at 06: 25 and resolved it at 06: 26.
- The confirmed cause was a service authentication-permission change preventing review submissions to GitHub's API. Reversion restored service, but failed reviews were not promised automatic reruns. Reconcile each PR.
- Save organization, repository, PR, trigger, failure time, HTTP status, request ID, and redacted logs. Stop repeated triggers without evidence.
- Use human reviewers or an existing independent review process temporarily, keeping required reviews and test gates.
- After recovery, request one minimal review on a non-critical PR. Check comments, completion, duplicates, and billing before gradually restoring automation.
- Fixed egress cannot repair a GitHub service incident. Check networking only when other requests also show DNS, TLS, proxy 407, or timeout failures.
Overall Operational status does not rule out a specific failure
The independent incident record starts at 04: 39 Beijing time on September 5, 2026. At 04: 57 GitHub confirmed failures for some code-review users and an identified cause. At 05: 54 it was applying mitigation, estimated about 30 minutes to recovery, and raised impact to major. At 06: 25 mitigation was complete and monitoring began; resolution followed at 06: 26.
GitHub later explained that changed authentication permissions prevented affected reviews from submitting to its API. Reverting that change restored normal operation at 22: 26 UTC on September 4, or 06: 26 Beijing time on September 5. This concerns service-side review submission, not a user's repository or personal-account permissions.
Resolved closes the service incident; it does not prove every review in the window completed or will rerun automatically. No automatic-retry guarantee was given. Preserve the incident ID, observation time, status, and impact, then reconcile affected PRs individually. Use the specific incident timeline even if the overall Copilot component is Operational.
First contain the issue: Freeze unrelated changes and retries
The confirmed fault concerns GitHub's review-submission authentication, not established missing permissions in your account. Changing proxies, CAs, organization model policies, repository rules, and accounts together can create a second failure. Freeze those variables and record organization, repository, PR, reviewer entry point, client, failure time, HTTP status, request ID, and redacted errors.
If automatic rules repeatedly request Copilot reviews, pause duplicate triggers or set a clear interval to avoid task buildup and additional AI credits. For an unknown result, inspect the PR timeline and review state before resending after a timeout.
Maintain review coverage with people and existing gates
A failed Copilot review must not lower merge requirements. Assign human reviewers and retain tests, static checks, required reviews, CODEOWNERS, and two-person review of high-risk changes. Delay only automated reviews that can safely wait.
Official documentation supports requesting Copilot from GitHub.com's Reviewers list or adding it through the REST API. Do not repeatedly trigger the same PR through several entry points during an incident. Choose one for recovery verification while keeping the human process elsewhere.
Verify with one non-critical PR after recovery
Once the official incident is monitoring or resolved, select a non-critical PR with a small diff, no sensitive data, and an existing human conclusion. Record the time and request one Copilot review. Check the overview, complete comments, task completion, and duplication of old comments.
The documentation says manual re-review must be requested again through Reviewers. Automatic re-review after new commits needs Review new pushes configured, and may repeat resolved or downvoted comments. Include duplicate comments and unexpected credit use in recovery checks.
Copilot's assessment is not the only approval
By default Copilot leaves Comment, not Approve or Request changes. Its approval assessment does not count toward required approvals. Even with public-preview Copilot approvals enabled, new commits invalidate previous approvals.
One reappearing comment is therefore insufficient proof of recovery. Check human approvals, required checks, branch protection, and expected reviewer states before restoring automation or merging critical PRs.
Separate service incidents, permissions, budgets, and networking
The incident covers service-side failure. Missing review access can instead reflect organization policy. Exhausting user, enterprise, or cost-center budgets can block AI-credit features. For an isolated 401 or 403, check permissions and budgets first.
Investigate networking when DNS, TLS, proxy 407, TCP timeouts, or other simultaneous GitHub/Copilot connection failures provide evidence. See the enterprise proxy and CA guide for proxy, Kerberos, and CA configuration. Changing IP does not explain away a confirmed service incident.
Completion criteria and when to fall back again
At minimum, require official monitoring/resolved status, a successful single review on the small PR, complete comments and overview, no duplicate side effects, intact human gates, and normal budgets and GitHub Actions minutes. Restore a few repositories first, then selected automatic rules, then wider coverage.
Pause automation and return to people if the incident reopens as investigating, failures recur, reviews hang, comments duplicate, billing is abnormal, or required-review state conflicts. With independent network evidence, visit PuppyIP for fixed-egress information. It can help reproduce a connection path, not fix GitHub's service.
Sources
Frequently Asked Questions
Was this a local configuration problem?
No local cause was confirmed. GitHub Status says service authentication-permission changes prevented API submission, with recovery after reversion at 06: 26 Beijing time. Do not change user accounts, proxies, certificates, or repository rules on that basis.
What caused it, and will failed reviews rerun automatically?
Changed GitHub service authentication permissions blocked affected submissions; GitHub reverted them. The incident page promises no automatic rerun. Reconcile PRs and request one normal re-review where needed.
Should I keep requesting reviews during the outage?
No. Inspect timelines and results first, then pause retries or increase intervals to avoid duplicate tasks, comments, and AI-credit charges.
Can I merge directly after a Copilot failure?
Do not lower requirements. Use human reviewers while retaining required reviews, tests, static checks, and two-person checks for high-risk changes.
How do I verify GitHub's recovery?
Request one review on a small, non-critical PR with an existing human conclusion. Check overview, comments, completion, duplicates, and billing.
Can the approval assessment replace required approval?
Not by default. Copilot usually leaves Comment; its assessment does not count toward required approvals. Copilot approvals remains public preview and depends on administrator settings.
Can a fixed IP fix this outage?
No. It cannot repair GitHub's service incident. Fixed egress has diagnostic value only when independent DNS, TLS, proxy 407, or timeout failures also affect other requests.