Currently, synaptic random number generators in BlueCellulab are initialized using the local cell id (node id within a population). Local ids are only unique within a population, whereas NEURON/ParallelContext and Neurodamus rely on a global gid (population-unique identifier obtained after population offsetting). BlueCellulab was therefore seeding synaptic RNG streams differently from Neurodamus. After switching the RNG seeding to use the global gid and aligning the population offset computation with Neurodamus, the voltage traces match between the two simulators. I will open a PR implementing this change.