The other issue is that the dual chip designs do have some chip to chip latency as well, so they’d be spending twice the vCache chips to produce a part that would likely perform worse in the thread light/cache heavy scenarios VCache sees the most uplift in.
Finally, the 5800X3D vCache really was a prototype product. The frequency limits have been confirmed to be from voltage constraints, not thermal limits, so I’d expect getting two of that generation of them playing nice on the same CPU was probably exponentially harder than doing only a single.