Please explain your conclusion, I don't see this creating a single point of failure, and it's quite the opposite.

A peering node would simply be another full instance of bitcoind running on a server that's well connected and very well resourced.  It would be able to service hundreds of connections from other, smaller, local nodes.  It's only purpose would be to ensure a well connected host closer to the smaller or home miners to hopefully increase their chances of getting their shares into the network faster.

In addition if it served the data that windpath is alluding to, and serve as a closer location to make the API calls to, it decentralizes the network further, and moves the processing of this data out closer to the individual miners.