On October 28, 2025, between 15:35-15:10 UTC (17:35-18:10 EET), customers using the Document Understanding OCR service in the West Europe region may have experienced processing delays or failures. The issue was caused by a significant reduction in processing capacity due to infrastructure node failures. The issue was automatically detected and resolved through automated recovery mechanisms. All services returned to normal operation by 18:10 EET. No data was lost during this incident.
To enhance our service's resilience, we're taking proactive steps to reserve additional compute resources. This will help ensure our Document Understanding OCR service remains reliable and performant under varying workloads. We're also collaborating with our service provider to investigate the root cause of the resource shortage and identify further measures to prevent similar incidents in the future.