It's too bad it "fixed" the bug.
The problem is that during an install a subvolume won't be created unless its volume is already mounted. But mountsCache won't believe that the parent volume is mounted unless it is given the proper subvolspec. When doing a BTRFS install, it appears that the btrfs format for the parent is given no subvolspec so failure during subvolume creation occurs. Essentially, I think subvolspec is assigned correctly during discovery, but not during creation.