The repetitive mode ensures that the same seed value (starting value for the random generator) is used. This way the same random values are drawn as the previous simulation run. If all the calculation are the same, the run will be exactly the same.
However, some user complain that the repetitive mode doesn't work. This is because the calculation are node the same. Often label, attribute or table values are used to perform a certain calculation (e.g. find out to which channel a product must be sent). If these values are not reset to the same starting value at the beginning of the run then the results of the simulation will be different. So if you want to re-run a simulation run always make sure that all stored values which are used in the calculation are reset.
Note that the seedvalue of the current run can be seen and set in the runcontrol. Another option is to use the 4D Script function GeneratorSeed(). With the 4D Script it is possible to store the SeedValue (e.g. in the tracer or a text file) of a simulation run so the results can be duplicated in the future.