An issue that came up in conversation with one of the Csound devs has got me thinking (thanks Steven!).
Each time we add a new CsoundUnity source we are creating another instance of Csound. With dozens of CsoundUnity sources it could get pretty CPU hungry. So I was thinking about creating a CsoundUnity child component. We could fill the child audio buffer with audio streamed from a master instance of CsoundUnity. That way we can use one Csound instance to generate audio for a range of components, and Unity would do the spatialisation for us