[NGL-Bi] Deals with readsets in case of resume pipeline

should we remove them ? Keep them ?

How to get the data base as clean as possible