For a limited period (12:00 CEST - 13:45 CEST), some content requests temporarily ran on a slower processing path instead of our optimized search system. This caused higher-than-normal response times and increased load for the affected requests.
We identified the cause — an internal configuration that, under certain timing conditions, kept requests on the slower path longer than intended — and have corrected it. As an immediate measure we restored the affected requests to the optimized path, and we've added a safeguard so the situation cannot recur. Performance is back to normal.