A human cannonball

There was an Air Force pilot named Mike Collins, a nephew of former Army Chief of Staff J. Lawton Collins. Mike Collins had undergone eleven weeks of combat training at Nellis Air Force Base, near Las Vegas, and in that eleven weeks twenty-two of his fellow trainees had died in accidents, which was an extraordinary rate of two per week. Then there was a test pilot, Bill Bridgeman. In 1952, when Bridgeman was flying at Edwards Air Force Base, sixty-two Air Force pilots died in the course of thirty-six weeks of training, an extraordinary rate of 1. 7 per week. Those figures were for fighter-pilot trainees only; they did not include the test pilots, Bridgeman's own confreres, who were dying quite regularly enough.
Extraordinary, to be sure; except that every veteran of flying small high-performance jets seemed to have experienced these bad strings.

In time, the Navy would compile statistics showing that for a career Navy pilot, i. e. , one who intended to keep flying for twenty years as Conrad did, there was a 23 percent probability that he would die in an aircraft accident. This did not even include combat deaths, since the military did not classify death in combat as accidental. Furthermore, there was a better than even chance, a 56 percent probability, to be exact, that at some point a career Navy pilot would have to eject from his aircraft and attempt to come down by parachute. In the era of jet fighters, ejection meant being exploded out of the cockpit by a nitroglycerine charge, like a human cannonball. The ejection itself was so hazardous—men lost knees, arms, and their lives on the rim of the cockpit or had the skin torn off their faces when they hit the "wall" of air outside—that many pilots chose to wrestle their aircraft to the ground rather than try it… and died that way instead.

Tom Wolfe, The right stuff.