While working to fix some caching issues yesterday, we inadvertently introduced a bug that caused erroneous cache collisions.

This issue may have affected your requests that had cache enabled. If you were using different metadata or cache namespaces, your requests were properly isolated and unaffected from this bug.

Please Note:
- The collisions did not happen between cache partitions
- The collisions did not happen across accounts

The issue has been fully reverted as of Jul 23, 2025 3:23 UTC.

Our team is actively monitoring the system to ensure stability.

We are sorry for this!