Cloud Run instances are being killed with OOM (HTTP 500) during traffic spikes that coincide with increased Firestore queries for product browsing. How do you prevent crashes and reduce Firestore load while optimizing performance?

Paginating with cursors and limits is the most effective solution. It retrieves only a subset of data per request, significantly reducing the memory footprint on Cloud Run instances and decreasing the read load on Firestore. Cursors ensure efficient pagination by continuing from the last retrieved document, avoiding re-scanning. Creating a custom Firestore index might improve query speed for specific fields but won't reduce the amount of data returned or the memory consumed if the query still attempts to fetch too many documents. Paginate with integer offsets is inefficient for large datasets as Firestore still processes all documents up to the offset, leading to higher latency and cost. Increasing Cloud Run memory allocation is a temporary fix for OOM errors but doesn't address the root cause of inefficient data retrieval and will lead to higher costs without optimizing performance.