What component allows data to survive a node crash in a cluster?

The most appropriate component that allows data to survive a node crash in a cluster is the Object Store. Object Store is a persistent storage solution in MuleSoft that can retain data even if a node or application crashes. This means that any data stored in the Object Store remains available and intact, allowing for seamless recovery and continuity when the node comes back online.

Hazelcast is often used in distributed systems for in-memory data storage and caching which provides quick access to data across nodes. However, it does not inherently ensure the persistence of data in the same way that the Object Store does, which is designed to allow data completion and retrieval post-node failures.

Using VM queues could facilitate messaging within a single application instance in a MuleSoft environment. However, they do not provide durability against crashes since their data resides in memory.

Round Robin, on the other hand, relates to load balancing and not data persistence. Thus, while understanding the role and functionality of each component is vital, the Object Store stands out as the only solution that guarantees data persistence across node failures.