Why is this interesting: First, it shows why diversity tends to collapse in small samples. By missing some of the producers of metabolites we can cross the tipping point. In the equations we can model this effect of sampling as an attack on the network: We replace M(x)->M(A(x)).