From approx 5:56AM UTC until approx. 6:07AM UTC, all API calls were timing out across Khepri. Engineers have traced the cause to a cache failover under unusually high load during a one-off data import. Failover took longer than usual due to the heavy load, and, combined with a recently added backoff strategy that was overly aggressive, the temporarily unavailable cache caused all available threads in the backend to stall. After the cache finished failing over, the system returned to normal operation.