We (batch mates) were discussing about time reversibility and Karan pointed out an example of system that is not in equilibrium but seems stable: all the particles in a box are moving back and forth in one direction perpendicular to a wall. Now this system is not in equilibrium but in a meta stable state. 

My question is that there are many-many such possible states and during the counting of the states of micro-canonical ensemble we add all such states. Even though such states can never be achieved by a system in equilibrium. Why even then we have correct results for observables ?