In terms of where this work should lead, there are many things that have to happen before any of this is ready for the ``real world''. Noise is always an issue when real sensors are used movement is imprecise. Tolerance for noise will have to be handled everywhere. Presumably that is the whole motivation for the section on the deviated Kissing Circles algorithm.
Perhaps rendezvous at the origin isn't the most appropriate model to work with. If, upon discovery of the target, one agent were to try to track down the other, using knowledge of the strategy being employed, then some of the overhead in rendezvous could be reduced in favour of a greater expected exploration payoff.
Asymmetric strategies should also be considered. Perhaps of the flavour of introducing a probability of attempting a rendezvous after some given iteration or just continuing to search with a complimentary probability. Thereby introducing the possibility of failed rendezvous attempts. An effect that will have to be tolerated any way under noisy circumstances.
Clearly the conjecture made in the last section needs to be affirmed. Hopefully though, this work is just the core for more to come.
Excelsior.