A recent discussion of this practice was included in an excellent Science Times section of The New York Times which focused on "learning what works" in education. The article looks at research from a team at the University of South Florida which looked at interleaved math instruction in a small sample (140 students in all) in a Tampa middle school. The students found that interleaved problem sets took a bit longer to do, at first, but that when they needed to study for a test they could spend less time reviewing because all the material they had learned was still fresh in their minds.
Furthermore, the results when the students took tests were remarkable. The scores on problems presented in the traditional manner averaged 38% -- but they scored an average of 72% on the material they covered through interleaved problems. The Times discussion included opinions of various psychologists and learning researchers as to why interleaving is beneficial, which ranged from noting that it engaged particular kinds of memory (something we call "paired associate memory"), to opining that since students need more time when they are first working with interleaving, the additional assistance they may get from their instructors makes a difference in how much they learn.
The research team plans to expand its investigation to see the impact of interleaving when it is used more broadly, with a far greater number of students. We will be interested to see the results.