As a lead developer for a retail system running on Cloud Run and Firestore in Datastore mode, you initially implemented an MVP feature that displays all available products by fetching the entire list from Firestore. After go-live, you observe Cloud Run instances crashing with HTTP 500 errors due to memory limits during peak traffic, correlated with spikes in Datastore entity reads. How can you optimize system performance to prevent Cloud Run crashes and reduce Datastore read operations?