It is quite easy to show the outward force of a cam is twice the pull.
However, for small pulls the strength of the springs can be significant, as can little rugosities in the rock which can be crushed out with higher forces. This partially explains the phenomena where a cam stays in for a test but blows at higher loads.
IIRC the assumed coefficient of friction on between the cam lobe and rock is ~0.3 meaning that it will support a sideways load of 0.3 times the perpendicular force. In the case of a cam this force is twice the load but you have a lobe each side so the cam will support ~0.3*2*2 times the pull on it or around 1.2 times the force, so you have something like a 20% margin.
If the rock-cam interface is 20% more "slippery" than this then you've got no margin.
Grippier cams can be made - for instance different materials or different camming angles but this comes at a penalty, weight, durability, range etc.
It is possible to make a lower range grippy cam for specialist applications in slippy rock (eg polished limestone) but I guess no one sees a market for this.
My own experiences of cam failures have always been in crap placements and the cams were always placed somewhat as a last resort.
The only injury I've had was a big cam in a crusty flaring break - the cam held a lot of tests but before moving I gave it one last tug and it failed smacking me in the teeth and nose and nearly causing a nasty fall - I now always lean to one side when testing big gear.
Post edited at 23:39