The correct option is You are billed for the provisioned RU/s regardless of consumption.

This is because Azure Cosmos DB charges based on the throughput you provision for a container or database. If you allocate a high number of RU/s you will incur costs for that provisioned capacity even when your workload does not consume all of those RUs. Overprovisioning therefore increases cost without guaranteed proportional gains in real world query latency or efficiency.

Provisioned RU/s do not affect billing or performance is incorrect because provisioned RU/s directly determine billing and they set the available capacity for operations which can affect performance when they are too low or unevenly distributed.

Cosmos DB will automatically decrease allocated RU/s when usage drops is incorrect because Cosmos DB does not automatically lower a manually provisioned RU/s. There is an autoscale mode that can adjust throughput within a configured range, but you still set the maximum throughput and billing is tied to the configured throughput model.

Query throughput automatically scales in line with the provisioned RU/s and improves performance proportionally is incorrect because increasing provisioned RU/s raises available capacity but query performance depends on indexing, partition key design, and the RU cost of the query. Higher RU/s does not guarantee a proportional improvement in query performance.

Full AWS Practitioner Certification Question