detector blinding is old news, and it is exactly the kind of thing people drag out when they want qkd to sound more exotic than it is. once you get past the optics demo, the weak spots are usually the boring trust assumptions around the box, the classical auth channel, and who is allowed to touch the fiber. so yes, there is qkd-specific testing here, but it is not some deep new class of math attack. it is mostly checking whether the deployment can survive an engineer, a vendor, or a network team being sloppy.