There are a couple of issues that should be addressed:
Sim# uses a non-standard method for generating normal distributed random variables, it's also somewhat slower because it produces one normal distributed random by drawing two uniformly distributed numbers. Instead it should use Box-Muller, Marsaglia polar or Ziggurat algorithms.
Generation of log-normals could be simplified by providing an API that allows specifying the desired mean and standard deviation and calculate the transformations
Sim# provides a FastRandom RNG that does not seem to be supported anymore, and the .Net system RNG. The family of permuted congruential generators (PCG) seems a promising alternative.
Sim# can make use of only a single random source. It would be beneficial to have multiple random sources in order to reduce variance.
There are a couple of issues that should be addressed:
FastRandom
RNG that does not seem to be supported anymore, and the .Net system RNG. The family of permuted congruential generators (PCG) seems a promising alternative.