I agree that certain roles, like warfare, are much less tolerant than others of individuals unexpectedly refusing to perform the functions of the role.
I agree that certain roles (again, like warfare) are much more likely than others to involve intermittent periods of extremely high risk, which makes steady-state reward very easy to game.
I agree that both of those factors seem to explain a lot more of what we call heroic than the "useful to the community, who can serve as models for others" metric proposed earlier.

It is not clear to me that we're less willing to tolerate astronaut's deaths or soldier's deaths than we are construction workers' deaths... we seem pretty willing to tolerate all of them, though we make a much bigger fuss about the former group. I suspect our lack of support for the space program has other causes.