Systems must be black boxed with sane single input single output modules... it is really unfair to attempt to poke holes in good computing foundations with the examples of problems that arise from sloppy computing.
This I can agree with but in the real world, most sound models are maintained by unsound IT staff.

Without hyperopia there is no begining... with myopia there is no end. The procress of creating any system must start from a global view and flow toward the goal with increasing percision. Focusing on single points with no context only leaves you with questions.
Would you still feel this way if you had people on the hyperopic side of the curve invisioning things that cannot be possibly done by those on the myopic end? For example, the project managers decide that they need a security solution that flys people across a pond, makes a pizza while in flight and then shines their shoes upon arrival on the other side. Oh yeah, and they need to be invisible, bullet proof and anonymous while in transit. Engineers are myopic (that includes myself) and as you pointed out, managers are hyperopic (which is where all the bad ideas begin and/or technically impossible ideas on a good day). Take my word for it, starting from hyperopic and flowing nicely to myopic is well and good in theory and in classrooms but in the real world, it floats about as well as a lead turd. There are many factors that influence design flows and without considering them, you yourself are myopic which contradicts your argument.

However, there are ways to hedge the deck and get a design flow somewhat close to your example. This is where technical project managers come in but there are very few good ones out there. Good ones have the ability to filter out all the white noise (politics, incompetence, budget restrictions, hardware limitations, security concerns, legal issues, ROI, etc.) before going forward with the design. When you shake out all of these things, you end up with requirements, the true starting point. I know you understand this.