[Engage] Incorrect timestamp format in Delta Share exports

Update timeline

1. identified Mar 16, 2026, 10:45 AM UTC

   During the weekend, data exported through Delta Share unintentionally used a different timestamp format. Timestamps in the exported Parquet files were stored as UNIX timestamps. Depending on your implementation, this may affect how timestamps are interpreted. We are currently rolling back the change to restore the previous format. No data has been lost, and we will run all exports again after rollback.

2. resolved Mar 17, 2026, 12:48 PM UTC

   This incident has been resolved.

3. postmortem Apr 01, 2026, 10:06 AM UTC

   ## Summary On March 15th 2026, an issue affecting Delta Share was identified. The issue was caused by a change in how timestamp values were stored. This change was introduced as part of an infrastructure upgrade on March 13, 2026. The incident was limited in scope and, based on our investigations, only one tenant/partner was affected. ## Customer impact One tenants/partners Delta Share integration failed when loading data containing timestamp fields. This meant that some shared data could not be processed correctly in their integration. At this time, no other Delta Share users have been identified as affected. ## Root cause As part of an upgrade, the infrastructure used to prepare shared data was moved to a newer cluster setup. Following this change, timestamp values were written in a different format than before. While the data itself remained correct and available, one partner integration was not able to handle the new timestamp format. ## Mitigation We worked directly with the partner that was affected by the issue after the issue was reported. The partner updated their integration to support the new timestamp format, which resolved the issue. We also evaluated whether to revert to the previous format, but decided not to do so since no other users appeared to be affected and a rollback would have required significant additional work for the impacted partner. ## Next steps To reduce the risk of similar issues in the future, we are taking the following actions: * Evaluating validation of data types when making infrastructure changes We are sorry for the inconvenience this caused.