I would argue that any network that's designed for interoperability will always have a vulnerability. Whether it's designed to be isolated or not isolated, ultimately it's going to have the same basis of vulnerability. I think Stuxnet proved that rather effectively in Iran, where a completely isolated system still managed to be compromised through a vector.
For me, coming at it from the point of view of security, it's that we need to change our mentality about how we think of security. It shouldn't be the Maginot Line, how to keep threats out. Rather, it's how you actually detect threats that you know and implicitly understand will exist within your network, and shape and manipulate them so as to minimize their effectiveness.