One point that I didn't have time to make was that high mortality is not generally evolutionarily advantageous. Even if you cooked up a strain that was especially nasty, it could take considerable effort to prevent it from mutating in the wild into something less so, since the mutated virus would have a survival advantage of not killing its host. Unlike machines, a creator can't really control what happens to a biological system in the wild as it interacts with the environment. In the scenario you describe, there is considerable uncertainty as to whether it would actually spread as engineered.
This is the crux of what I've clearly done a poor job of saying: we know so little that all these scenarios still rely on incredible amounts of luck more than technology.
Not terribly reassuring :) Perhaps 1918 wasn't a very high-fitness virus, but it still got ~5% of us before burning out.
That leaves the threat level >> (conventional) terrorism.
If you're not trying to create a "supervirus," but just shotgun a series of viruses to create maximum health disruption, then (as you know!) it's already within our technical grasp. Expensive, time-consuming, and probably failing 90% of the time, but accessible nonetheless. Considering how bad our hit rate is on the trivalents, I could see a plague of multiple, high-lethality strains with a wide variety of hemagglutinin and neuramidase antigenic shifts as a plausible bioterror scenario.
It's not a doomsday scenario, since we have such good reporting and analysis infrastructure, and much of the heavy engineering on the problem (like alternate vaccine production methods) was done during the earlier bird flu outbreaks, but it would be a very nasty kind of terror attack.