Given everyone would have different models, the compilation process (including placing and routing) would absolutely destroy things. I've seen them take 1+ hour

Given everyone would have different models, the compilation process (including placing and routing) would absolutely destroy things. I've seen them take 1+ hours on a fairly fast machine. The problem with connecting N logic blocks is that you need N^2 buses to connect them directly to one another. Luckily they aren't but that's why the placing/routing optimization problem takes so long to compute as transistor count is now in the billions.