Customer Impact

Between 09:41 UTC and 13:24 UTC, customers using IXP in the U.S. experienced generally poor performance in the form of slow API requests.

In addition, these specific actions had increased error rates:

• Consuming stream
• Uploading documents

For a small number of specific clients, streams were unable to be consumed entirely during this time.

Root Cause

An IXP database migration took place in the U.S. region at 09:41 UTC on February 3, 2026.

This migration changed the underlying database queries used by several of our API routes. One of these database queries was poorly optimized, leading to bad performance when executed against the high volume of data present in the U.S. region. This database query was used by the streams and document upload routes.

Detection

Our engineering team was closely monitoring the migration, and observed an increase in stream response times. In addition, automated alerts picked up on the increased error rates and automatically alerted us while an investigation was underway.

Response

The team worked to find and fix the problem with the database query. At 13:04 UTC, a hotfix was deployed. By 13:24 UTC, API response times and IXP performance had returned to normal.

Follow-Up

We recognize how disruptive this incident was for our customers, especially those relying on timely document processing and stream analytics. The IXP team will ensure that in future, similar migrations:

• Have better performance testing
• Have faster rollback capabilities