Carabiners are incredibly strong—they meet a minimum test of 20 kN (4496 lbf or 2039 kg) when properly loaded their major axis with the gate closed. In an open gate scenario, carabiners still test to a minimum of 7 kN (1574 lbf or 714 kg). But when you test a nose-hooked carabiner, it can fail at less than 10% of its rated closed gate strength - that's less han 2 kN (500lbf or 227 kg), a load that can be easily generated in even the smallest of climbing falls or even just a light bounce test.
WHY?
Why is the carabiner's breaking strength so low when loaded in this manner? It's a combination of an open gate scenario coupled with the fact that the carabiner basket is being cantilevered off of the bolt hanger/sling/Stopper wire, meaning the load is not in line with major axis (i.e., the carabiner's spine). This off-axis loading causes the carabiner to be excessively torqued and break at an extremely low load.
BD TESTING
Black Diamond manufactures a lot of carabiners, and therefore Black Diamond tests a lot of carabiners. We not only understand the loads at which carabiners break, but also the modes (i.e., location of breakages), depending on the way it was loaded. So it's possible to look at where a carabiner is broken and have a good idea of how it was loaded.
The photos below show typical failure locations for one style of carabiner tested in four different configurations. As you can see, a nose-hooked carabiner will most often break at the top of the spine, while open and closed gate failures typically occur at the bottom of the spine, and minor axis failures almost always occur at the gate.
[Disclaimer: All carabiners are different, and detailed analysis of the particular carabiner's geometry and failure modes is necessary in order to be able to estimate the particular loading scenario with any level of confidence.]