Now that we have multiple implementations of the metadata API, they may support different variations of parameters. For example, the dataset-based metadata storage does not support cursors in conjunction with relevance(default) sorting. The Elasticsearch implementation, however, can support that.
Right now the getValidatedSearchRequest() method in the MetadataHttpHandler hard-wires the constraints for the dataset-based implementation, and we have unit tests that verify that.
In the future, this validation should be delegated to the metadata SPI, and test cases should not expect a particular implementation's behavior.