There is the idea that there is no time in a completely closed (thus unobservable) system. Within such a system, a subsystem may be imagined to be split off by some virtual boundary. However, one wants to continue to consider the remaining system as if it was the complete system (which of course is not correct). In that case the description of the effects of the subsystem on the remaining system requires "time".
Does anyone know more about this idea?
