So the topography here is: Individual router satellite mesh net SpaceX router connected to the internet. Is there any reason why the protocol between satellites matters? They clearly aren't aiming for interoperability with other companies.
Each satellite will have basically three different RF setups on it:
1) L or S-band TT&C (tracking, telemetry and control). This is the admin and OOB management for the satellite and some of the frequencies can be seen in SpaceX's recent FCC filing, with earth stations in the Seattle and Los Angeles area.
2) Multiple Ka/V-band or other multiple spot beam antennas aimed straight down, serving CPEs.
3) There will be at least two dedicated high capacity dish antennas per satellite aimed at large, high capacity fixed earth stations (2.4 to 5 meter size) operating in the Ka-band or higher. The earth stations will need to be geographically distributed around the globe at various fiber IX points, similar to how o3b has done things now, so that the satellite-to-satellite trunk links can take the fewest possible number of hops before downlinking to the terrestrial internet. The more medium sized earth stations the better, because it increases the chance that a link to an individual CPE can take a bent-pipe path rather than traveling through multiple satellite-to-satellite hops. For example a CPE in far rural Wyoming, and a satellite passing over the general area of Wyoming, might be able to "see" a 3.0m sized high capacity earth station located near fiber in Cheyenne. Data from CPE, up to satellite, down to Cheyenne, one hop.