Design a Scalable Like Counter

Design a large-scale service that stores and serves like counts for content items. Requirements: - Users can like and unlike an item. - Product pages must display the current like count with low latency. - The system should handle very high write volume, potentially billions of like events per day. - Duplicate requests and retries must not corrupt the count. - The design should separate the source-of-truth interaction data from a read-optimized counter store. - Changes should be propagated between stores using a change-data-capture or event-streaming approach. - The system should handle hot items, recovery, backfills, and consistency trade-offs. Describe the APIs, data model, write path, read path, storage choices, scaling strategy, and failure handling.