to answer your last question first: the simulation duration is defined in LTE_config.simulation_time_tti where one TTI corresponds to 1ms.
For your first question, it depends on your simulation setup (and thus is only implicitly visible from the documentation). Assuming static users, their location and therefore the large scale path loss doesn’t change over time. Thus, only the small scale fading/channel realisation for each link (desired as well as interfering) will vary. For moving users, also the large scale path loss will change.
There are more aspects that are varying over time, e.g., scheduling or traffic models, but this should already give you an idea for a basic simulation.
If you have more questions on this topic, please let me know.
the only way how to do that in the current version of the simulator is by averaging over several simulations, where you define/create new user locations for each run. You could also try to define your own walking model where you let the users jump to a random position after a fixed number of TTIs, but I am not sure if the extra effort outweighs simply performing several simulations with new user locations. (Also then you couldn’t change the number of UEs during one simulation).