There is a long-standing TODO in simulations.py that might affect the performance of simulations. It would be better to parallelize the call to generate_spin2ecl_quaternions when MPI is being used.
(See this comment by @marcobortolami in issue #184.)