It shouldn't take 10 years, but it isn't a simple trip into a hackerspace either.
To put it in the familiar context: what you're describing (and what DoD is used to) is a mainfraime. These guys just made a PC.
You're right, my post wasn't a list of requirements, just an illustration of some of the things one might end up having to care about when packaging up a military product used in dangerous situations. There's a balance to be made, it should be made on a case by case basis.
This is precisely part of the mechanism that causes every military to be well prepared for the previous war but unprepared for the current one.
I propose greater emphasis actual feedback mechanisms based on experience in the field. These can be made to work on timescales of months or weeks. When bureaucracy comes into things, iteration timescales can stretch to years and change can become generational.
I get frustrated too at too-slow, too-big companies running projects with billion dollar price tags resulting in no workable product, but I'm not as sure that the solutions are simple. I am sure that without constant oversight from within and without it could be much worse.
Google's approach to building server farms was likewise to buy hardware that was ultra-cheap because it had failed to meet manufacturing specs. Google expected a certain percentage of hard disks, etc., to fail, and simply swapped them out when they did. That resulted in a significantly lower TCO. (So says Steven Levy in in his recent book In the Plex.)
You might be able to tweak specifics to account for this, but you also might not.
No way terrorist types could misuse that one.
Also, highly-adapted droid != highly-adaptable droid. There is no need for an amphibious droid in Afghanistan, just like one doesn't use the same apparel in all climate zones. As long as interface/principles are the same (like a PC), various versions of it can be used in (almost) any environment without extensive re-training.
On the bright side, at least they weren't forbidden from using there own solution.
I am not allowed to comment about the specific project I have worked on, but I can tell you that a trip to the hackerspace would have resulted in a better product in that respect too.
We should all count our blessings that hackers tend to be peaceful people by nature. Imagine if we were trying to kill each other.
You can build some really neat things when you have a military R&D budget behind you. The downside is of course that it actively or passively hurt people.
And then there's the whole calculus-of-death/unintended consequence thing, where in theory, a more accurate missile is better for everyone because it reduces collateral damage. But because it kills less innocents, there are fewer qualms about using it more often. This applies strongly to the current UAV trend - You can politically spin (or just bury in accounting) "Ooops, We lost a $xMM drone", versus actual pilots dying.
This reminded me of a very similar quote from Edward Luttwak about the disconnect between students of war and the people who fight wars: http://www.youtube.com/watch?v=7vz1BpRf3Lc&feature=playe...
The entire interview is fascinating and worth the time. It's a bit dated (from 2007), but still has a number of really interesting tidbits throughout.
Edit: Although I guess I can extend that to human societies in general, not just hackers.
Edit 2: Man, no idea who manages to disagree with that.
I have been working for a defense project about instrumenting some armored vehicles with video cameras and monitors. The project was more than 10 years old when I arrived in it. I couldn't help but thinking "bring the prototype to any hackerspace and pay 20k of salary for two weeks of dev and you'll be done."