A couple things I'd pull out:
Program managers sit limited terms to prevent empire building
Program managers have a great deal of autonomy once funding is initially allocated
DARPA has a high appetite for risk. They're ok with 90% of projects failing to hit their goal.
