Since we have bitcoin nodes all over the globe it would be pretty easy to connect to them to create the block templates, so anyone running a pool node can connect to his own node or nodes near him to get a list of pending transactions/current block and etc.
Every pool owner should be running their own node and not rely on anyone. Otherwise, it is a recipe for disaster. I believe you know that.
There is no need for a supernode as far as mining is concerned, however, mining is location-sensitive, unlike with Bisq latency needs to be pretty low, so pool nodes need to be located across the globe for this to work effectively.
Bisq was an example of a DAO that works in practice. You can remove mandatory Tor from the equation, use location-based algorithms and you have improved latency significantly.
I don't know if we're envisioning the same thing. The point of this DAO would be to elect representatives which will be responsible for broadcasting the block headers to the miners. As far as I understand, by design, PPS requires miners to agree on a single block header. That can't happen if the network is completely peer-to-peer; there needs to be an entity which will broadcast the unique block header, and the rest of the miners will only have to verify it comes from that entity. Representatives will be the main operators, as pool owners are currently in centralized pools. Governance proposals can be later on made by any stakeholder.