I'm wondering if software defined radio has anything to contribute here. Ultra wideband could help to mask the radio signals. Software defined antenna's could help with the directionality problem. This stuff is expensive today, but maybe not in a few years..
I'm really surprised that P2P doesn't scale on a mesh and would like to understand this better. I do research on message passing algorithms and obviously trees (hierarchies) are great, meshes are not... I can see that the overhead of routing messages is going to grow (like n^2?) with the mesh size, but i'm surprised there is no way around this.
In general, the market has overtaken hobbyist hacking by a wide margin. Any technique you are considering has already been considered and either adopted or abandoned by the Broadcoms of the world. Basically all radios are now SDR-ish but their firmware only contains a highly-optimized implementation of a single protocol (e.g. 802.11). Sure, a spinal code PHY with a mesh MAC will beat 802.11 by some percentage, but it costs 10x more because it's an FPGA instead of an ASIC so you end up switching back to Atheros/Broadcom radios.
I'm really surprised that P2P doesn't scale on a mesh and would like to understand this better. I do research on message passing algorithms and obviously trees (hierarchies) are great, meshes are not... I can see that the overhead of routing messages is going to grow (like n^2?) with the mesh size, but i'm surprised there is no way around this.