Thanks, those links were very elucidating. Point being that carabiners can fail and they may not show any indication of an imminent failure during an equipment check — ...”when the carabiner is un-loaded, all surface cracks completely close”... This makes a catastrophic failure of a “perfectly good” carabiner a possible result after enough cycles. They do have a life expectancy.

The Ringling Bros failure in the second link was a demonstration on how to induce a carabiner failure. The ‘biner that failed was stressed in three directions, but not along the major axis.
... “The cause of the failure of the carabiner was the manner in which it was loaded, subjecting the carabiner to tri-axial loading in violation of industry practice and the instructions of the manufacturer. The carabiners are designed to be loaded in their major axes along the spine.”...

It totally makes sense now why the “D” shaped carabiner — the stress is shifted closer to the spine, reducing stress on the gate side.