So I pulled out my 2nd anchor, which was a CQR and it worked right away. I have no idea why the Rocna would not stay put ... maybe it has problems if the ground is not level?

Possibly you've found one of the (few?) niches where the Rocna won't do :-) I suppose it's relatively dependent on on gravity to set correctly (but not as much as the spade) and hence may be dragged sideways if the bottom isn't level.

In the end of the day, when it comes to anchors, it's more about religion than science.

For the causal anchoring situation, what is really relevant is ... what is the probability that the anchor won't set or will start dragging if dropping it completely on random on a place where we don't know the bottom conditions? One of the "selling point" with Rocna is that it's supposed to be an all-round-anchor, it should not be needed to know much about the bottom conditions, and one should not need to change the anchor dependent on the bottom conditions ... but here you've apparently found a place where the CQR excels compared to Rocna.

For the causal anchoring situation it's not really relevant weather the anchor will lose its grip under Beaufort 10 or Beaufort 12. Very few people would leave the boat by anchor under such conditions anyway (though, I do want to have the theoretical possibility to survive such storms by anchor, and I do tend to believe it's a good idea to let the wind hit the bow under such circumstances, rather than letting the wind come from the side while staying in a harbour)

Tests i.e. done by boating magazines won't really cut it, they drop the anchors maybe some dozen times under controlled conditions and check how much force the anchor can take. However, to really check the relevant metric it's needed to drop the anchor a thousand times, on all kind of different bottom surfaces, and make statistics based on that.

User reports are also kind of useless. All together, there is sufficient statistics in the user experiences, but they are hard to collect. Many owners will boast "I've never dragged with anchor X, hence anchor X is really good" - but it doesn't really matter if the captain is rarely using the anchor, and only using it in very calm conditions on well-known anchorages. The captain has only tried anchor X, but it's quite probable that anchor Y, Z or Q also would have done the job. What really matters are the user experiences where the user has been experiencing dragging.

Quite often one would read on forums that a captain has switched from Bruce to Rocna due to problems with dragging, and the captain has experienced that the Rocna was really superior ... but for one thing, what is untold is that while upgrading from Bruce to Rocna, the size of the anchor was also upgraded. For the other thing, the captain probably doesn't realize it himself, but he is biased - he believes the Rocna is better, and he has invested quite a lot in the Rocna. He wants the Rocna to be better than the Bruce. Under such circumstances human nature tends to overlook all observations not supporting the belief.

Despite those problems, I still believe that I've done a pretty decent job sieving through the information I could find, and I'm still convinced that the Rocna (and by extention, my Mantus) is a better all-round-anchor than the CQR. Perhaps I've become ... religious :-)