Continuous stream processing with low latency is the correct option.

Continuous stream processing with low latency is designed for unbounded data where there is no defined start or finish and it processes events as they arrive which keeps end to end latency low.

The approach supports event by event processing, windowing, and stateful operations which are required when you need near real time insights from a continuous flow of data.

Microbatch processing for frequent scheduled jobs is not suitable because it relies on fixed schedules and it introduces latency that is higher than what low latency streaming requires.

Cloud Dataflow is a specific managed service that can implement streaming workloads but it is not the class of workload itself and the question asked for the type of analytics workload rather than a product.

Traditional ETL batch pipeline is built for bounded datasets with a clear start and finish and it cannot provide the low latency needed for continuous unbounded streams.

Full AWS Practitioner Certification Question