Well, I'm hesitant to talk about it too much, since I don't want to generate any unrealistic expectations. Since wme2 will be rebuilt from the ground up, in the first phase it would be nice if it supported roughly the same feature set as wme1 (through the years wme got to support everything and a kitchen sink...).
Of course, there will be many improvements to remove weak points of wme.
The most obvious is a new renderer, based on Ogre3D. This brings support for Ogre .mesh format for models, i.e. a more comfortable content creation pipeline for many 3D modelling applications (.X format is a PITA). Also there is a robust material system, replacing the .fx files used by wme1.9.
The scripting engine is rewritten. Although the language stays approximately the same (similar to ECMA Script), some annoying limitations and bugs have been removed.
The engine is written with portability in mind, unlike wme1 (Ogre is a major factor in this).
The tools should be completely redesigned, however I haven't touched that at all yet.
Regarding 2.5D / 3D scenes, both are to be supported eventually, although I don't know if both in the initial release (those two systems are quite different, perhaps more that I initially thought). Plain 2D scenes are just a simplified case of 2.5D.
And of course, tons of small things.
I gathered you're working on a new engine as well...? How is your progress/goals?