So far as I can see, PyLag advects the particles and stores their positions in memory until it's finished the simulation. It'd be nice to save the outputs periodically so that if PyLag crashes (!), we have at least got the simulation outputs up until the last dump to netCDF.