In the large research area of wireless body area networks, cooperative applications involving several users is attracting strong interests. This cooperation may target a simple information exchange or even some cooperative decision such as swarm coordination. We consider in this paper such a swarm of users moving in a common direction and we are interested in the mechanisms allowing to propagate and share some common information. We extend and improve a previous algorithm derived as a max-consensus approach. We describe a complete experimental setup deployed during a real bike race with 200 runners. %We additionally deployed a spying code allowing to store the complete information flows on We provide a deep analysis of the radio links properties (stability, strength...) and the equivalent graph evolution (graph diameter, neighborhood characterization...) from the record made on 20 nodes. At the best of our knowledge, this paper is the first one providing such analysis in a real experiment with a swarm of 200 nodes located on bikes during an official race.