The Plugin API indeed is about Exporting, Editing, Opening, and Publishing.
So if you wanted the pixel data, you would create an "IEditingPlugin" and you could get the rasterized pixels of the image. It would not hook into the Capture One RAW pipeline - which I think makes sense since different parts of the RAW processing happen at different times and it could be funky to know WHERE in the pipeline to make your pixel changes. This pretty much works like "Edit With..." and opening in Photoshop or something, but here you can specify the tool to run.
As far as tools to modify parameters in the UI - it seems that Mac users can use AppleScript to control some of those properties. Not sure how deep that goes though, as I am no longer on Mac.
So in the example of doing PixelShift - you could totally do that, but with an "IOpenWithPlugin" so that you have the RAW file with unmosaic data. Then you would return a result - which would be the new file with your changes that Capture One can handle, probably DNG. I am thinking about writing a plugin with
https://www.fastrawviewer.com/SonyPixelShift2DNG, once it's out of beta. At least that's the theory glancing at the API docs.
All that said, I would love more power APIs - like being able to modify Keywords (have AI/Machine Learning tag my photos), or even modify Mask data. And I certainly would welcome UI APIs like suggested so Windows users have the abilities that Mac users have with AppleScript.