Another lesser-known feature is Spark's support for approximate queries. Instead of computing exact results, Spark can provide quick and approximate answers for aggregate queries, which can be extremely useful for large datasets where precision is not critical.

Finally, Spark has built-in support for vectorized UDFs (User-Defined Functions), which can significantly speed up the execution of certain data transformations. This feature leverages hardware acceleration to process data in batches rather than individually, resulting in improved performance.

One lesser-known feature of Spark is the ability to define custom partitioners. This allows developers to have fine-grained control over how data is distributed across the cluster, which can greatly improve performance in certain scenarios.

Spark also provides support for user-defined accumulators. These are mutable variables that can be updated in a distributed manner. This feature is especially helpful for tasks like collecting statistics or monitoring progress across the cluster.