Eh, the problem is that any such project should be tied to efficiency. And it should be able to ramp up very slowly, and should be cancelable any time. Usually the problem is that these projects run into issues that need escalation, eg. a small change in scope, requirements, specs, budget, structure, milestones. They usually have to start by building bridges to other groups, and there's usually no mechanism to make good incremental agreements, that are needed to facilitate a great systems design. (So in the end a lot of thing hinges on random/arbitrary interpersonal relations and chance.)

Just delivering something for the sake of it leads to atrociously bad public infrastructure, be it software or a stretch of pavement.