Resolve collision-label scope and flooding-rate dependence
Determine whether the collision-label discrepancy caused by counting network-wide resource overlaps while the analytical collision probability considers only in-range collisions, and characterize how the reference implementation’s collision probability varies with attacker rate and population by sweeping those parameters.
References
Two questions stay open. Our collision label counts an overlap anywhere in the network while Equation~eq:pcol resolves collisions in range alone, so the two are not measured over the same set. And the reference implementation's collision probability rises with attacker rate, by $0.339$ at 200 packets per second and $0.393$ at 1000 at the sparsest density, while we evaluate the surrogate at the higher rate alone. Sweeping attacker rate and population would settle both.