I was slightly torn on the usage of the word "parallelise", in that yes, CPU-wise, no tasks will progress in parallel.
However, the point of the library is to allow certain parts of a pipeline to progress at the same time in the real-world. In the example, they are calls to asyncio.sleep, but in real cases, they could be HTTP requests. As in, the bytes of the request/response will be going across the wire at the same time, so the total wall-clock time of (certain) pipelines will go down.
So, I claim "parallelise" _is_ appropriate.