Phrak, you haven't addressed my question yet: what happens when there are momentum carrying fields involved?
If \sum_i\textbf{p}_i represents the sum of the momenta of all particles in a closed system, it needn't be constant, since fields can transfer momentum to particles. Momentum for the system will still be conserved, but only when you include the momentum in the fields (which will be an integral of a momentum density). However, when fields are present (even if you include their sources in your system), the sum of the forces on all the particles in a system need not by zero, even if momentum is conserved .
Consider a field propagating at finite speed (like light) towards a stationary particle. Prior to the interaction of the field and the particle, there are no forces acting on either and the only momentum is stored in the field. After the two interact, some momentum may be transferred to the particle, and hence some force was exerted on it. Momentum is conserved in this interaction, but force is not since there is no equal and opposite force exerted on the field (which is presumed massless).