Ideally this would be the case. In an ideal world, the same seed leads
to the same random events and thus to the same simulation results.
However, there are also parts in the simulator which are not under
control of the development team, i.e. especially in the network layer.
There, other sources of randomness might be used leading to a totally
different path of random events.
So it may be, that by using the GNP or modular network layer this
happens. However, with a static simple network model this should not
happen (although then the results are not so valid).