I'm not really in a position to care to argue for this (dropped the project months ago), but for the sake of discussion:

1.) the internet already is decentralized ... decided to shut off every piece of equipment they had today it
would cause a problem for about a week

Hardly - We pay ISPs in order to use their infrastructure - and we abide by their rules or we don't get access (remember Net Neutrality?). I don't see how you think if ISPs shut down there would still be access. Without their centralized infrastructure, there is no network.

2.) each node would add like 20ms of latency to any connection

Where did you get this number? The overhead of hops in modern wireless mesh networks are routinely under 3ms. With distributed data (guaranteed small packets) 1-2ms overhead would be likely per jump, and with increasing capacity of wireless technology, that means less hops and more reliable wireless transmission.

3.) your proposal looks overly complex

It differs from B.A.T.M.A.N. and all those other systems in that it is distributed data - this is why it looks more complex, because it breaks data up and distributes it in parallel across the network - this increases security and speed at an exponential rate that scales with the number of users.

4.) ISP that state they are not allowed to resell the connection

This isn't about reselling connections - it would require licensing unused wireless bands (most likely wiMAX frequency areas) which is difficult, and one of the reasons I'm not pursuing the idea.

5.) "trusted nodes" on the network are going to need to be able to encrypt data at wire speeds

There wouldn't be any trusted nodes, or encryption - the data is split up (again, distributed data transfer) so that at any local computer relaying data, all they see is a piece of the original request. Maybe they get a password if they are monitoring their through, but without the username it's meaningless. This is why it's faster and more secure to use distributed data transfer, rather than a single, closest to destination route like every other mesh network project I've seen. The only way someone could intercept a request is through a massive, expensive physical infrastructure that spans miles and intercepts each and every piece of a request - hardly worth the time of a hacker looking to steal passwords or credit card numbers.

6.) That's no going to be possible. Unless you put it in a different universe someone is going to move data from one to the other, and then everyone is going to do it.

A "normal" server can't propagate or read the data of this type - servers have to transmit and receive data in parallel across multiple transceivers, piecing the resultant pieces together to form the request. And regardless, it isn't the data that is going to be so worthy of attention, it will be the network itself. This isn't just about being "anonymous" it's about faster speeds, and LEGAL freedom from regulation.

7.) The existence of a free network

For as long as we need energy to support the networks infrastructure, there will never be a truly free cost internet