A little appreciated fact is that the internal forces generated in cams placed in downward-flaring placements can be arbitrarily high. See, for example, Vaino Kodas's
explanation. Among other things, he says, "...as the angle of the flare approaches the spiral angle the forces grow very large (theoretically infinite)..."
What this means in practical terms is that cams in flaring placements are susceptible to forces that will break either the placement or the cam (or both). I'd guess that usually the placement breaks in some way and the cam is extracted by the fall, but breakage is another possible failure mode. This is a consequence of the camming geometry and is not a design failure.
This is one of the key posts in this thread. Anyone who has been reading this thread and who hasn't already clicked through to the Kodas paper, ought to do so. For those too lazy, his conclusion on cams in flares reads as follows:
"A little experimentation shows that as the angle of the flare approaches the spiral angle (tan-1µ) the forces grow very large (theoretically infinite). There will be a large compression force on the cams, a large shear force on the axle, and a large spreading force on the walls of the crack. Most likely, at these high forces, significant bending and/or deformation would take place so that the above equations would no longer be valid."
In other words, forces on cams in flared placements are much higher than in parallel-sided placements, and we ought to expect "deformation." One could reasonably expect "material failure" in even a normal cam in such a placement given a high-load fall. In a fall in which the cam was leveraged, which was clearly the case in this instance, forces would be higher and more complicated.
The Kodas finding is significant because its predictions proved out in the informal testing that Malcolm has described in which a variety of brand name cams, in off-plumb placements, failed in ways that includedwhat we've all been rather vaguely referring to as "material failure." Or rather, the inference that Malcolm, Richard, Michael and other knowlegeable folks are making is that the placement begins to shift or fail and then weird forces multiply and stuff starts to break.
I find it difficult to imagine a generalizable lab situation-- i.e., one that could become an industry standard test --that could really usefully replicate the wild variety of flare/angle/lean/leverage situations we find the the big wide world.
We've all probably taken at least a few falls onto weird cam placements that we knew to be sub-standard, but to judge from what I read on this site, the collective willingness of folks in the community to take those sort of risks has actually increased. Or maybe the ability to recognize those risks has diminished. Or both. My guess is that as more and more climbers are willing to regularly lob onto cams in weird flared placements, we are going to see more and more "exploding cams," and not just those of the "my-tweakeremployee-forgot-to-braze-it" variety.