I voted MoGraph, just because I believe Blender is severely lacking in that category. But the best part about Blender is that it can do all these things relatively well
That and the fact that specializing Blender on one thing would betray one of its primary purposes as outlined early on in its life, and that is to give anyone access to modern 3D technology with few to no barriers (whether that would mean a financial one or a technological one).
To get a non-restricted general 3D app. that is clearly superior to Blender would require the spending of more than 1000 Dollars or Euros, which is quite a massive leap to make should Blender stop being a general app. (that and the fact it enables the ability to stay away from DAZ and their penchant for prioritizing their 3D dolls over the actual development of their software).
I think that Blender can be the perfect tool for indie game development, used with Unreal or Unity (or others)
There is too much work to maintain and improve BGE to the level of these “free” alternatives, so I wouldn’t focus on it, but on game assets creation (PBR viewport, texture painting, sculpting improvements, export tools, outliner / asset management…).
It is some kind of funny to read all these votes. Blender has generic lacks and optional lacks. First generic problems should be solved (viewport performance, lack of Max-like modifier operation, removing current edit mode (proper multi-edit mode instead), etc.), then it should be examined what areas should be improved to get more money into the development (like PBR viewport or MoGraph features).
Focus on those things all users are using first.
Start with modelling and UV’ing. Be it a game or movie you always need models with textures.
Blender could be the best modeller, its already better than Maya and Cinema.
Throw in a node based procedural workflow like Houdini and it really starts to get interesting.
It seems to be a legacy of the past when computers were weaker. Many times I need to edit multiple meshes in the same time, UV edit multiple meshes at the same time.
For both there are addons, but mesh multiedit never worked properly for me (many bugs, bad output), multi UV edit is slow like hell.
Also it is pretty illogical to use features both in Object mode (like bevel) and edit mode; in this I prefer Max approach far more (and if I remember correctly, MODO will work/works this way, too). It is a great thing we have modifier stack in Blender, but to be honest I almost never use them as it makes no sense to use them with current implementation where edit mode and object mode are separated).
When I came to Blender a few years ago, community gave many entertaining responses, too, based on the experience (or lack) they had (no offense). Since then on my criticized areas many things were fixed or became a feature through addons and the community told these should have worked this way from the start. So I do not really care about community responses (do not forget, about 90% of Blender users are hobbyists, no offense).
The situation is that anything what makes the life of Blender users easier is a good thing, but development have to focus on generic areas where everybody is able to profit from.
Also I could write lot’s of areas to improve:
Outliner improvements
Layer management as it works in other apps
Increasing viewport performance
GPU selection for rendering (also improving OpenCL integrations, etc.)
Custom menus/icons without coding skills (I feel so sorry that Menu Pinning addon is not in development anymore)