Mesh networking isn't implausible at all, hams do it constantly. That is exactly what a packet TNC does. The real question is can they do it in such a small device and with such small power requirements, but 100 meters is nothing for a modern digital mode even at low power levels, so I don't consider that unrealistic either. In fact, I'm hoping that they are under stating the effective range, because 100 meters isn't practical except in some pretty dense urban areas. I'd say the bear minimum effective ranges start at 300meters. And nor is communications sans (commerical) infrastructure implausible, for that is the very nature of mesh networking anyway.
Packet TNCs are set up at fixed locations and use manually-configured routing though, right?
AX25 networking protocol would largely be manually configured, but that's no longer a requirement. They were never really intended to be fixed. Although more modern modes don't call themselves packet radio and don't use hardware TNC and don't use AX25 protocol, they have their basis in packet.
That's not mesh networking in the sense we're talking about here. The hard part of this kind of mesh networking (I believe the technical term is "mobile ad-hoc networking") is dynamically-updated routing between a bunch of small moving devices, all of which has too little memory and compute power and bandwidth to store a full global routing table. Normal mesh networking is similar but has stationary nodes that are typically more powerful.
A dynamicly updated routing table, or a routing table at all, is not a likely condition of this kind of network. Sensor style networks use passive monitoring to detect reachable peers on the fly, and it's a completely different model than a network that is trying to emulate the mass data transfer capabilities of a wired network. A bitcoin transaction doesn't need to be routed, it simply needs to be broadcast.
