QA, along with their own testing, gets input from the customers (who aren't always reasonable...'what do you mean, this software WON'T wash my car and cure male pattern baldness? YOUR salesguy said it would!' .... and feeds it back to R&D.
R&D looks at the input. A third of the time they say "Uh.... yeah, this is software, not a 'car wash' And I dunno about this baldness thing." A third of the time they say "Yeah. That doesn't fit into the current roadmap." (translation: 'This is hard, and I don't feel like doing it.'). The final third of the time, the process works pretty well.
It's fun to watch, from a distance. I'd not want to be in either place, though. :-)
Compound all of the misery you mention when the customer is the U.S. Government.
Then compound it again when the the customer’s office is filled with your own cast-offs.