I'm wondering if I should reinvent myself as a Plato scholar. I find every year that the part of my courses where I'm talking about his writing ends up being my favorite to teach, largely because it's where I'm most likely to discover something myself while preparing to cover the same material yet again.
So, current working theory of Republic, which may not be original but I don't think is completely orthodox either; the point of the book is an extended attack on what Nietzsche called the master morality.
To review, the master morality is the morality of Homer's epics, an invention of a proud aristocratic class. According to the master morality, to be good is to be like the aristocrats, or at least like their self-image; strong, brave, honest, cunning, rich, and good looking. Generally able to succeed in life, to get respect and to get your own way. Badness is everything that is unlike the way of the masters.
So why think this is the target of Republic? Well, here are a couple of interesting bits of evidence. Very early on, of course, Cephalus voices the opinion that only rich people can afford to be honest, one of the central tenets of the master morality. Somewhat later, Socrates says, and Polemarchus agrees, that the theory that justice is the ability to help friends and harm enemies was surely invented by some very powerful man, a Xerxes or the like. Thrasymachus, of course, identifies justice with whatever advances the aims of the ruling class. And the complaint of Glaucon and Adeimantus that justice is not advantageous to the just person seems to be based on thinking that it won't contribute to having the ideal aristocratic life.
Further, there's the whole of book X, which I feel is widely misunderstood. It is commonly taken to be an attack on all art, but Plato picks Homer as his example over and over again in this book. If I'm right about the master morality being the target, this Homer-bashing is of course exactly what we'd expect.
I'm sure there are aspects of the data my theory doesn't fit, but it seems to me at the moment that it works really well. If so, perhaps I should try to write a paper on it, assuming nobody else has already done so.