In my experience (and I do do development in environments where this comes up frequently), it is not at all unusual for applications to either rely on buggy behavior of another piece of code, or to make unwarranted assumptions about how another piece of code works, that happens to be valid under some circumstances but not others.
Which is why Linus' policy is that, when the choice is between userland compatibility and better adherence to theoretical documentation of the kernel interface, tie goes to userland compatibility. This is smart - Linux already has enough issues with userland apps breaking backwards compatibility to scratch itches that nobody really has (see GNOME 3, the entire audio stack, etc.); if the kernel is disciplined, at least, it keeps the whole environment from turning into a bug-ridden tar pit.
As for Linus' attitude, well, I have to agree with Linus on this one. Fix the mistake first, either by removing the patch that broke everything or quickly implementing a fix, then ask questions. The first rule of kernel development should always be "Do No Harm". If you're in charge of some part of kernel development and you find yourself breaking that rule, you need to un-break it ASAP, then assess how you found yourself breaking it in the first place. Unfortunately, Mauro wasn't grasping that - he was too busy asking "reasonable questions" to undo the damage that his commit did when his first priority should have been to stabilize the kernel. Don't get me wrong, Mauro's questions are important and they do need to be answered, but only after userland is back to the condition he found it before the commit. And not until then. And *certainly* not with the mainline kernel.
I agree with the deadwood issue, but there are also some dynamics that favor having work done by government. The big one is that there's essentially no profit motive. In a well-functioning federal agency, all of the staff are encouraged to "do the right thing" for the people they serve, rather than maximize profit.
The real problem with the lack of profit motive isn't feather-bedding or anything crazy like that, though a fair amount of that goes on. It happens in the private sector, too. The real issue is that management is kept accountable by profitability - if they get penny-wise and pound foolish, the market will (eventually) punish them for their shortsightedness (yes, yes, after Wall Street's computers spike share prices for a month so investment bankers can extract every ounce of value out of the company first). In government, however, the only thing keeping anyone accountable is popular opinion. If the people think you're saving money, even if it's by throwing together 23 different layers of "accountability" between a funding request and the request being fulfilled, they'll reward you. If the people think you're spending too much, regardless of value you're returning to the community, your funding will be cut. The result is that pay and funding are directly tied to appearance of performance (plus bits and pieces of patronage, where available), not to actual performance or value.
The worst part is when the same people running government institutions decide to treat their systems as "best practices" and insist on forcing private companies to adhere to them, too. Then you end up with well-meaning but ill conceived bits of legislation like Sarbannes-Oxley.
Say "twenty-three-skiddoo" to logout.