Something interesting about these is that while unit to unit variation is extreme, each unit's drivers (L and R) are generally matched to that unit's FR. For example, both drivers in pair #6 have the same severe dip at 15hz. And both driers in pair #3 have a severe dip at 175hz and another at 7k, with the same wild peak characteristics. That seems like considerable evidence that Beyer knows how much variation there is between units, because they are measuring drivers well enough to match them between units that have wildly different FR's. Unless their system is completely automated, but it seems like someone must be there supervising that process. This has to be some of the craziest driver matching ever.
it's worth noting that the pair with the worst most wonkiest FR, has the best driver matching. It's pretty funny if you think about it.
You could call this something like micro-QC, macro-FAIL.