The SAMPLER should write json file on the fly (update) and not only at the end. Hence, if the simulation does not make it to the end, we have still have a dataset that can be used.