We investigate cross-layer optimization to route information across distributed wireless body-to-body networks, based on real-life experimental measurements. At the network layer, the best possible route is selected according to channel state information (e.g., expected transmission count, hop count) from the physical layer. Two types of dynamic routing are applied: shortest path routing (SPR), and cooperative multi-path routing (CMR) associated with selection combining. An open-access experimental dataset incorporating 'everyday' mixed-activities is used for analyzing and comparing the cross-layer optimization with different wireless sensor network protocols (i.e., ORPL, LOADng). Negligible packet error rate is achieved by applying CMR and SPR techniques with reasonably sensitive receivers. Moreover, at 10% outage probability, CMR gains up to 8, 7, and 6 dB improvements over ORPL, SPR, and LOADng, respectively. We show that CMR achieves the highest throughput (packets/s) while providing acceptable amount of average end-to-end delay (47.5 ms), at -100 dBm receive sensitivity. The use of alternate paths in CMR reduces retransmissions and increases packet success rate, which significantly reduces the maximum amount of end-to-end delay and energy consumption for CMR with respect to other protocols. It is also shown that the combined channel gains across SPR and CMR are gamma and Rician distributed, correspondingly.