# Microsoft publishes review of Azure OpenAI Sweden Central outage

Facts read 16 Nov 2026. Page reviewed 16 Nov 2026.

Microsoft's Azure post incident review for 29 September 2026 says a planned deployment cleared an internal service's caches and sent backend traffic to about 20 times normal. Azure OpenAI Service and related services in Sweden Central had intermittent failures from 10:03 to 15:58 UTC.

The review says two other defects in how the internal service read from its caches caused most lookups to bypass the cache and call the backend service directly. Connection capacity on infrastructure nodes was then exhausted.

Microsoft states that the extended duration was primarily caused by health checks that depended on connecting to the database layer, which put instances into a restart loop. Service availability began to recover at 15:40 UTC after the health check was reconfigured.

The review lists repairs including throttling and rate limiting for the backend service, more caching capacity and better diagnostic metrics. It describes an availability incident.

## Related pages

- [Astrana Health files Item 1.05 after social engineering attempts](/news/astrana-health-social-engineering/)
- [Nutex Health says the third party published the data](/news/nutex-health-data-published/)
- [Boston Scientific says its August incident is likely material](/news/boston-scientific-material-impact/)
- [All of them](/news/)

## Sources

- https://azure.status.microsoft/en-us/status/history/ (read 16 Nov 2026)

Last reviewed 16 Nov 2026. Canonical: https://cloudsecuritybreaches.com/news/azure-openai-sweden-central-pir/
