(7/?)
We *already have* a way to interact with BS accounts without touching any part of the BS service. It's called BridgyFed;
Any time, money and effort we might spend setting up ATProto infrastructure would arguably be better spent shoring up Bridgy. Or setting up more instances of the code, so it's no longer a Single Point of Failure.
Because if @cwebber's analysis is correct - and I've yet to see it debunked - this approach would give us way more bang for buck.