A large bank in the European Union faced performance issues with its Neo4j cluster due to an increasingly large database, prompting a decision to scale out across multiple clusters using Neo4j Fabric. This approach allowed for sharding of data by time window, making it suitable for analytical workloads and batch operations. By distributing workload across smaller shards, the bank achieved significant performance boosts, easier maintenance, reduced infrastructure costs, and improved scalability. However, it's essential to consider optimal sharding strategies upfront, as different situations require tailored approaches, such as sharding by logical domain entity or geographical location.