Aerospike allows applications to process query results as a stream of records, which can be paginated, partitioned for parallelism, and resumed at a later time. Query execution is executed in a scatter-gather fashion, with the client library sending requests to multiple cluster nodes in parallel and gathering responses. The application can read and process all or partial records from the stream, save cursor states, and resume processing later. Queries can be run in sync or async modes, with the latter using a registered callback class to handle each record. Partition queries provide greater control over flow and parallelism, while pagination allows for retrieving results in smaller chunks until all are returned. The query results do not follow any significant order, requiring applications to sort records within themselves.