Not seen any mention yet about the ammunition source.
Are we talking about handloads?
If we are, I'd suggest that the charge and the barrel's harmonic node could be out of sync. This is one source of double grouping. If the load charge weight is altered up or down by a mild degree, does the outcome change with it? This can still exist with factory ammo, but well crafted handloads will generally have a smaller size for the two individual groups due to more precisely equalized charges. If factory ammo, try a different one, any different one. If it's ammo, the double grouping ought to change.
Another would be bedding (in a bolt gun). But before we dismiss bedding, let's stop to consider how the effect of a loose relationship between the Upper, Lower, and Pivot/Disassembly Pins just might mimic the effect of loose bedding. Don't forget to check for sideways slop, too; it could well be linked to this posted issue. This also relates to any motion in the optics mount and/or a loose reticle. The relationships between rifle groups under recoil need to be consistent. Slop translates into different barrel attitudes at the time of bullet emergence. It doesn't take much when one considers that the difference between a 100yd zero and a 1000yd zero is roughly 30 minutes, or 1/2 of a degree. If any slop between Upper and lower can be detected, there's a problem, and may be one of the key reasons why otherwise precisely built AR's may not shoot as tight as logic might suggest.
Understand that the aim of the rifle at the time of bullet emergence is always different from what exists when the sear breaks. The physics of recoil under third law principles insists that this is so. That is why component fit/integrity, consistent shooting position and grip, and other factors become so critical once we have already nailed down the mechanics of load development and production, and rifle configuration. When we can't identify a problem, are we also taking a moment for looking in the mirror? By definition, any problems humans can have are human problems, in some way or another.
Let's also consider Parallax, if the cheek weld is not clearly defined and the parallax adjustment is off because the parallax adjustment and objective focus are combined at the factory, but not in good sync for that particular scope (this is a lot more common than many might think, and is one of the reasons why expensive scopes cost so much more). This might be improved upon by verifying parallax adjustment with the head nod/wag test. If the reticle tracks across the target image in close conjunction with head repositioning, parallax is off. It may be necessary to choose between a sharp focus and a null parallax.
Does it occur consistently with different magazines?
I'm still backing LowLIght regarding shooter input; just 'cause several folks do it similarly does not always mean it's not a people issue. Do they all have a good LOP match? A bad LOP match will very likely generate problems on the target.
Greg