[zmgsabst]

I took a slightly liberal interpretation to their comment:

There’s two references to an underlying object, one of which presents the inverse of the other.

Think of it as there being a single strand of string - in a U shape. If I measure one end, I’ll measure the inverse twist direction on the other. Even if the twist direction is randomly chosen when I measure… because it’s a single string.

But only if I don’t have extra twists introduced along the U (ie, interactions with the environment).