Ok folks, I did some playing tonight with various carabiners connecting my rappel rack to my harness' D-ring. One thing that I learned, is that it's almost impossible to "break" the carabiner when the rack is in a 'normal' orientation.
I keep the 'open' side of my rack facing the right, and with the brake strand of the rope running to my right side. As shown, with a Petal Sm'd screw-lock carabiner. (Not my regular BTW)
Keeping the rack oriented this way, it's almost impossible to get a carabiner to cross-load. I'm fact, I tried several different carabiners, and this Petzl Sam's might have been the worst. If I was careful, I could get the rack's eye to 'snag' against the screw-lock, and with *just* the right pressure, I could get it to stay cross-loaded as I pulled up on the rack. (Simulating sitting down in the harness, and loading the rack.) However, cross-loaded like this, it still is holding the gate "closed" and is still within the rated cross-loaded configuration of the carabiner. As shown.
What I did learn after watching several YouTube videos, is that it seems only possible to break, or "lever" the carabiners to failure, if the rack is unweighed, and then rotated 180* from it's "normal" orientation. That means in my case, the rack would be rotated so that the 'open' end is facing the left, with the brake-strand of the rope facing the left too. See screenshot from YouTube video. It's hard to see, but his rack is rotated 180* from normal - open end facing his left...
(Ironically, this is the same creator of the "Swaygo" packs that I was discussing only a few days ago in another thread.)
And me duplicating this "failure" with the rack rotated 180* - open end facing my left. I could probably load the rack with my weight, and brake that carabiner - but only after rotating the rack 180* from normal...
Now, my regular linking carabiner is a DMM Ceros. It's anti-criss-loading feature seems to prevent even this from occuring. It's possible to rotate the rack, but pressure is kept off the gate. See below.
I legitimately tried to "break" this one by contorting and twisting the rack and carabiner every which way. But either the carabiner would self-correct, or bind/pinch in such a way, this no critical component was stressed.
Shown in the normal & correct orientation...
I'm summary, I'm quite happy to understand the why & how that carabiners can be broken with racks. It seems that it takes quite a bit more cocking-about with the rack, to create this type of leveraged, broken-gate failure. And knowing that, I'm quite happy to continue using the DMM Ceros. It's very secure, and quick to remove, without needing a wrench or pliers that a Mallion needs.