This is a minor performance optimization, but with 2.6 it should be possible to start using new methods introduced in 2.5 which will allow further streamlining some of common parsing tasks by BeanDeserializer. With bit of testing performance gain seems to be ~3%, i.e. measurable but limited.
Still, it's an easy win considering that most other opportunities have been exhausted.