Clay does a great job of highlighting the problem, but I'm disappointed that his solution is just an updated version of what got us where we are.
The problem is that we are trying to regulate a very complex and constantly changing system with a giant set of rules. You can think of government as a giant computer system that is constantly being modified. Something like procurement regulations is a module that affects almost every part of that system, and each procurement rule is a little hard coded rule in the module that looks something like: if cost > 250k then paperwork(x,y,z).
We know what happens in systems like this, you either can't make modifications in ways that make sense because you have to work around rules that weren't designed for them, or you have to keep changing the rules. Either way, you eventually end up with a giant pile of crap that doesn't work and no one understands, i.e. procurement law.
I don't think the solution is to keep adding new rules nor do I think we should restart the rule making process over from the beginning (sort of what +Clay Johnson
proposes). I think the solution is to come up with a better rule making process. So what would that look like?
Well, I don't know, but I would look to a few things for inspiration. Machine learning gives us very elegant solutions to complex problems that we used to try to solve with lots of hand crafted rules, if we can make a machine that beats Ken Jennings at Jeopardy surely we can make one that does a better job of legislating than Congress (I'm only half joking).
Another elegant solution, the one libertarians always bring up, is the free market. Unfortunately free markets often do a very poor job of producing the goods and services government provides (hence the existence of government), but that doesn't mean we can't use it for inspiration. Where free enterprise works, one of the secrets of its success is individual accountability. Unfortunately that is something that is often lacking at both the political and civilian employee level.