The overwhelming majority of observations I look at are live animals, so I would only ever use it for dead or indeterminate cases (which are probably only a few percent of the total I have ever looked at).
It depends on which groups you look at most, I suppose. Determining the sex can be a very important part of identification, particularly in groups where sexual dimorphism is very common (for example, Odonata). I would estimate that I use the sex annotation for at least a third of the observations I look at, but have so far never needed to use the alive/dead annotation.