The conventional explanation for metamorphosis is that it evolved gradually, with the juvenile form becoming specialised for feeding and the adult for mating, until they barely resembled each other. [Donald Williamson, formerly of the University of Liverpool] thinks otherwise. He points out that marine larvae have five basic forms and can be organised into a family tree based on shared characteristics. Yet this tree bears no relationship to the family tree of adults: near-identical larvae often give rise to adults from different lineages, while some closely related adults have utterly unrelated larvae...-- from Why Darwin was wrong about the tree of life 
...His star witness is the starfish Luidia sarsi, which starts life as a small larva with a tiny starfish inside. As the larva grows, the starfish migrates to the outside and when the larva settles on the seabed, they separate. This is perfectly normal for starfish, but in Luidia something remarkable then happens. Instead of degenerating, the larva swims off and lives for several months as an independent animal.
 (added 23 Jan): Larry Moran suggests a better headline for the article would be: "More evidence that Charles Darwin didn't know everything there is to be known about evolution when he published his book in 1859." (Added 9 Feb) Carl Zimmer on Crunching the Data for the Tree of Life.