Just read this piece by Richard Elmore and Elizabeth City from the May/June 2007 Harvard Education Letter, which does a great job of describing how the school improvement process proceeds in fits and starts. Elmore's work is intriguing because, unlike other research centered on how schools are organized, his work is based on observed teaching practice.  The authors explain why, as schools improve, there are setbacks:

We see this last pattern frequently when teachers go from asking students questions to which there is a correct answer to asking questions for which there are multiple possible answers. At first, teachers aren't very good at asking the questions or setting up a classroom environment in which ambiguity and intellectual risk-taking are valued, and students aren't very good at providing answers that require sentences rather than two-word responses, or at offering rationales for their answers.

Changing teaching practice takes time, but most accountability systems are not designed to distinguish between schools that are working hard and making progress and those that, frankly, need a kick in the pants.  The Mississippi school profiled in this WaPo article seems to be the latter, but even in this discouraging case, the school has a committed principal who would like to turnaround the school. 

It's a messy business and, while Elmore and City are critical of our current accountability systems, they provide little guidance on how such systems might improve. When NCLB reauthorization really begins to move, we will need to hear more from those folks who, like AFT members, have been in the trenches, doing this difficult work.