Julian Karch

@jakob-r thanks for pointing me to this indeed very relevant discussion. Seems the best approach, for now, is to either use another package (for example, clustermq) or manually chunk many...