Cache Tag Invalidation Limits in Multi-Server Clusters

TYPO3 employs a hierarchical caching system where content elements are tracked via cache tags to automate invalidation across the site. In a single-server environment, the "Clear cache" function effectively flushes these dependencies to ensure frontend consistency.

When scaling to a multi-server cluster using a shared cache backend like Redis or Memcached, the synchronization between the database state and the distributed cache entries becomes critical. There is uncertainty regarding how the system handles partial clears when different nodes may have varying levels of synchronization with the central cache store.

• Does the cache tag invalidation mechanism guarantee atomic updates across all nodes in a clustered environment?
• What are the documented limits of the "Flush cached messages" process when managing stale results across distributed nodes?