Kevin Padhian in Nature:
The answer, we think, is that closely related lineages have similar genetic components, so under comparable ecological conditions they are likely to produce similar mutations that are then selected for. Many call this convergence; I prefer the term parallelism for closely related lineages. 'Convergence' is appropriate for reinvention in very different groups — the superficially similar wings of birds and pterosaurs, or the elongated grub-seeking fingers of the aye-aye (Daubentonia madagascariensis) and striped possum (Dactylopsila trivirgata). We can catalogue examples all day, but is there any real theory of convergence? We cannot assert that some lineages are 'fated' to converge on these features. Biological ideas of determinism went out with Jean-Baptiste Lamarck in the late eighteenth century. Evidence against determinism is the prevalence of creatures whose adaptations have never been duplicated: the kangaroo, the platypus (Ornithorhynchus anatinus), the century plant (Agave americana, which blooms only once in its multidecade life) — and humans. Primates have the equipment apparently needed to evolve flight: arboreal habits, large brains, good coordination and active metabolisms. Yet no primate seems to have evolved gliding, let alone powered flight. No evolutionary duplication is inevitable.
More here.