Couchbase Server's CAP behavior varies significantly between single cluster deployments and multi-cluster deployments using Cross Data Center Replication (XDCR). In a single cluster setup, Couchbase is primarily a CP system, with the capability to prioritize availability over consistency through auto-failover mechanisms. However, when employing XDCR across multiple clusters, Couchbase shifts towards an AP system, offering eventual consistency by resolving write conflicts and ensuring availability across clusters. The document emphasizes that while CAP is a useful introductory concept for understanding system behavior, it may not fully capture the complexities of Couchbase's deployment topologies. It also provides a detailed table outlining the CAP balance across different deployment configurations, highlighting the importance of understanding XDCR's role in high availability and disaster recovery scenarios.