The point of the project is to exercise the API, not necessarily to support a specific subset of the existing plugins; we would want a variety of inputs, outputs, filters, and codecs to ensure that each of those APIs was well-exercised. Selecting a variety of pre-existing plugins gives us a pretty good idea of the types of configurations plugins expose to the Logstash Pipeline Configuration format, which should help shape the needs of the API.
There is a Work-In-Progress PR, along with a lot of discussion -- you can take a look at the current approach to build context.
And, because the point of the project is to shape the API, we want to exercise it as much as possible; if the API makes writing plugins in alternate first-class JVM languages unreasonably difficult or awkward, that information may help us to iterate on the API in a way that makes it less awkward overall. Writing the exact same plugin against the API in multiple languages could be a great way to find where it needs to improve