We have a few problems right now that need fundamental changes to how patents work in order to be resolved:
Firstly, there is a flood of patents far too great to allow patent examiners to examine each patent in detail. We can't solve this by adding more examiners; there's no money for that. We can't solve this by allowing an arbitrary backlog; sooner or later we'd get to a point where you'd spend longer for your application to be processed than the patent would last once approved, which would hurt legitimately useful applications. The current solution, just doing less work per patent, just means that more junk patents come through.
Additionally, we don't have enough experts. A patent on "storing a word processor document in a single XML file" (real patent) might not sound obvious to a patent examiner who doesn't have a deep understanding of IT but to an IT professional it's blindingly obvious; after all XML is a universal format and we store all sorts of other documents in XML form already. Still, a patent has been granted for this "innovation", most likely because the patent office can't afford enough IT experts to properly evaluate every IT patent. (Admittedly, the patent is specific enough that one can, with effort, create a non-infringing XML text document format. But it's still obvious.)
Of course it doesn't help that some granted patents are overly generic. Many patents just declare dominion over an idea, sometimes even without providing technical information on how to make the idea actually work. This can be hard to see for the examiner because of the relative dearth of domain experts.
Compounding that is the fact that willful infringement nets harsher punishment. However, if I actually do the research to make sure I don't violate certain patents it becomes reasonable to assume that I know about all relevant patents in the field. If I overlooked some and end up infringing them it becomes difficult to prove that I didn't know about them, costing me more money. Thus, the safest course of action is to never read any patents at all so I can at least claim ignorance. This keeps me open to surprise litigation, of course, and it also perverts the entire point of the patent system: Patents are not there so that someone can control an idea, they are there so that someone provides his idea and technical work to everyone else in exchange for some royalties.
Fixing this mess won't be easy. We need far more experts, more time per patent and fewer patent applications. The former two aren't going to happen because nobody's willing to pay that much money and the latter isn't going to happen as long as obtaining patents is as lucrative as it is today. While I don't think that killing off the entire patent system is the way to go it's easy to see how people come up with the idea.