It is unlikely that such a disease would evolve naturally; too much genetic diversity in the species; for any massively fatal disease there will be a naturally immune population, and it will die out due to lack of hosts too rapidly to infect sufficiently widespread anyway.
This doesn't hold for tailored diseases, and one of today's nightmares is that you don't need a Manhattan project and billions of dollars to develop a new lifeform; it could be done in the laboratory of a hospital somewhere, or a fairly advanced school. Certainly, to know exactly what a particular virus will do, enormous amounts of computer power, but if all you're trying for is kill as many as possible then a simple gene splicing set up, a test population and a reasonable amount of luck could be enough.
The disease should have a long, infectious but symptom-free incubation period before a short lethal period. This bypasses quarantines, and doesn't give laboratories enough time alone with the organism to find its Achilles heel before the patient dies (or the lab techs).
The infection vector needs to be simple - no secondary 'the bite of a Tasmanian hobbit shrew' which has become the favoured pet of the rich – respiratory or skin contact have the best wide distribution record, and who are we to argue with success?
Remember, a naturally occurring disease that invariably kills its hosts is a failure. After all, its primary imperative is the same as yours or mine; feed, reproduce and try and make sure your offspring have as much chance of survival as possible (under which analysis I'm as much a dead end as bubonic plague; but I like to think slightly better liked). A really successful disease hardly kills anybody, and is only a short step away from symbiotic. Most of the really lethal infections have crossed over from other species, where they had a much lower death rate.
A synthetic organism has no such limitations. I'd suggest reading "Infectress" (by Cool) or Herbert's "White plague". It is born to kill, and targeted; programmed homicidal.