We realised that existing pods were rendered ineffective due to an internal issue that caused a steep increase in connections and repeated Redis timeout errors. We think the impact from the issue is over.