Also, this thread may make it seem like I'm against single points of failure.
I'm not. I think SPOFs are great.
I think more things should be built with SPOFs. "Sorry guys, power's out at my house, you'll have to use someone else's [THING] tonight"
The lights don't always have to turn on, the weather isn't always going to be good, flexibility is the key to happiness and success.
Human scale software, anti-capitalist software, is going to be full of SPOFs.
But, if we do it right, it'll also be federated and distributed such that any single point of failure, any node, any provider, can die and the network will heal around it and keep going.