What is the effect of changing the random seed? Which seed is closest to 'real life'?

The random seed influences many aspects of the model. It merely changes the start values of the random value generators used internally in the model. These values influence the arrival times of vehicle in the network, the stochastical variability of their driving behavior and also selection of a certain distribution value wherever distributions are used (e.g. dwell times, speeds, colors etc.). There is no seed that replicates 'real life' better than another. It's more comparable to the daily changes of the traffic patterns at the same location.