I've just finished Lee Smolin's book The Trouble with Physics. The book covers the recent (1980-2005) history of theoretical physics, with emphasis on string theory, and discusses some of the reasons for the lack of progress in fundamental physics during that period.
The problem is not thatthere have been no advances, per se. The problem is that there has been a disconnect between theory and experiment. String theory has been the dominant approach since the mid 1980s, but during that time it has never made a prediction that was both new and testable. The currently fashionable explanation for why string theory does not make testable predictions is that our universe is just one possible universe in a landscape of 10500 universes. That's a very big number: consider that the observable universe has "only" ~1080 protons. If we were to envision each proton as a mini-universe containing that many mini-protons, and repeat this construction for six iterations, we would still be short by a factor of 100 quintillion of the number of possible universes in the landscape.
This is science?! It sounds at least as much like a sick and twisted version of the infamous apocryphal final exam question, "Define the Universe. Give three examples."
The problems Smolin identifies are not unique to physics. As he correctly notes, the peer review system is well equipped for making incremental advances in science, but it is designed in such a way that its mistakes will be conservative mistakes. The next Einstein is unlikely to hold a tenure track faculty position at the time he makes his key advances in fundamental theory. For that matter, the real Einstein did not, either: he was a patent clerk in Switzerland in 1905, when he published the three papers that shook the foundations of physics.
My own field of space physics has long been suffering from a disconnect between theory and experiment. But at least there are doable experiments that can falsify some of the theories, and one of the leading drivers of mission selection is the need to address the predictions of theorists.