Capacity increased. Error rates returned to normal.

We are monitoring and following up with changes to prevent recurrence.

Between August 23 and August 31, during several periods, some requests to our offers endpoints timed out or returned incomplete results, affecting a subset of customers.

This was caused by a capacity limit in an internal service that evaluates offer eligibility. We have increased that capacity and verified the limit is no longer reached at peak traffic.

This incident is resolved.

Between August 14 and August 25, 2026, a subset of requests to Kard's Eligibility API returned errors or timed out during several brief windows, the longest on August 23–24 (UTC).

The cause was a supporting policy-evaluation service that exhausted its allocated memory under peak load, causing it to restart and briefly stop responding. Requests that did not hit the affected instances were served normally, and no data was lost or incorrectly processed.

We have increased the service's capacity and adjusted its autoscaling configuration. The service has been stable since. We are also making changes to the component responsible for the memory growth to prevent recurrence.