GHOST processed_slitflat() calls self.imaging_flat(processed=False...) - the processed=False appears to be a work around imaging_flat calling .PROCESSED_FLAT() on the query, but this has the undesired side effect of setting howmany=10.
It works, but it's returning (and thus DRAGONS is fetching) 10 processed_slitflats each time, when it only needs one.