If you want to see the concepts/problems/limitations of local optima in action play freedink game.(http://www.freedink.org). The spatial movement of the enemy agents get stuck in a local optima all the time. And that makes killing them easier.You just have to find the right position to get them stuck and keep hitting with ranged attacks. Also it is very easy to see they have a breaker if the agent repeats the same coordinates some set times.(will have to go through the code base to extract those). But all in all it is a good example of seeing the neccessity for improving AI algorithms in action.
P.S: Ofcourse with multiple agents in a screen it gets complicated and less likely to get stuck.
Update: Though in multiple agents, it’s rather easy to get them to hit them each other.. It seems to work sometimes and sometimes not.. i.e sometimes the attack doesn’t cause damage..interesting..