I think, perhaps, that the problem with terrorism is that the distribution of terrorism event impacts is likely heavy-tailed -- i.e., the probability of one "black swan" event (nuclear, biological, chemical) is higher than a normal distribution model of event impacts would suggest.
But, to me, this means that the national security apparatus should be exclusively focused on reducing the probability of those most-terrible events, and instead leave the small-scale stuff to standard law enforcement. I.e., if the # of deaths is almost certainly < 200 (say), normal law enforcement procedures apply. If more than 500, then every resource can and should be brought to bear to prevent it.
But, to me, this means that the national security apparatus should be exclusively focused on reducing the probability of those most-terrible events, and instead leave the small-scale stuff to standard law enforcement. I.e., if the # of deaths is almost certainly < 200 (say), normal law enforcement procedures apply. If more than 500, then every resource can and should be brought to bear to prevent it.