You manage a Cosmos DB container for an e-commerce catalog. Traffic is highly variable: it sits near 800 RU/s for most of the day but spikes to roughly 9,000 RU/s for two to three unpredictable hours during flash sales. The team wants to avoid throttling during spikes while minimizing cost during idle periods, and they do not want to manually adjust throughput or write scaling code. Which throughput configuration should you choose?