+1 on the suggestion of a headless (mostly) Matrix-6/1000 styled module, but with the suggestion that it should be expandable by adding (different) voice cards–and polytimbral.
Also - why not the idea of heterogenous voice-card modules, a la the BeagleBone cape EEPROMs? Each card type could identify itself to the system as Prophet, OB, or Mopho cards, while providing a means for the less well-heeled user to add voices gradually, or with a mind to unusual configurations (3 OB, 4 Prophet, 1 Mopho, for example), similar to a modular setup.
The way I see it - shared architecture pays for itself over many product cycles, so why not go wider with the Prophet / OB voice card platform?
I work professionally as a design engineering consultant / software architect, so my perspective might be a little different than others...years ago (middle 90s), long before I worked in an engineering capacity, I had the opportunity to visit Ensoniq in Malvern PA as part of a Guitar Center-sponsored group for sales + product training.
We all sat down in a conference room, at which point they asked us about products we'd like to see. I suggested, at the time, that they focus their efforts on modular, shared-platform products, a la a sampling drum machine, keyboard workstation, and digital effects processing engine, using shared PCB modules–all of which would allow them to build once and reuse many across multiple product types, including the burgeoning PC sound card market.
While it might be daft to suggest that ignoring that approach hastened their downfall (after all, they were building new, dedicated keyboard controller PCBs for EACH synthesizer generation, so who knows WHAT they were thinking!), I really think that smaller companies ought to invest at the front end of a design cycle, then churn out products that utilize the same LEGOs in appropriate combinations. But hey - what do I know?