It's possible (it's been done forever as "hybrid crossfire") but you'd lose some synergies of having the CPU and GPU on same die, unless you build a really expensive and fat cache coherent bus between the two (and even then latencies would be higher)
Right. I think the biggest issue would be the bus between the two chips.
Maybe they could do something like an APU with 1-2 CU's dedicated for compute tasks plus a full GPU so the full GPU isn't used for compute tasks. But I'm not sure that makes sense, just add CPU mores.
If there were dedicated chips for CPU and GPU, I think you could get more bang for the buck per chip, but the rest of the system would become more complex and expensive. I really think this would be driven my how expensive 7nm will be.