Caching high-latency network requests