hey Justin: so, to be clear, here's where we are. Algorand's network architecture follows a 2 tier model with nodes focused on consensus communicating through nodes whose main job is to relay messages. These nodes used to be permissioned (the old "relay" nodes). With the release of the new P2P architecture, we removed the permissioned constraint. Anyone can now create these nodes - we call them repeaters, and validators can auto-discover them and automatically connect to them. This eliminates the reliance one any permissioned element and establishes completely permissionless transmission routes. Validators have full control over how they connect to repeaters. They can decide to connect exclusively to permissionless repeaters or follow a hybrid model and connect to both permissionless repeaters and permissioned repeaters (the old "relay" nodes). It's up to them to decide. In our view, this allows for the network to gradually migrate from the old model to the new model, while preserving the ability for anyone to run a validator from their basement on a $300 computer with 1 algo in their account - the gold standard to be censorship resistant.