I haven't heard of any projects like that. Not for FBX. Not for Collada.
The longer story:
I think one reason why JSON was the chosen export format was so people who need really specific formats will have an easier time parsing and converting it for their purposes.
I can't officially speak for Esoteric, but it's always seemed like exporting to other Spine-independent formats is not a priority.
I think it's expected because other formats are 3D formats and inherently don't support the full set of Spine's features anyway. This would cause it to behave sub-par, and that's just no good. The same is true for other skeletal systems.
Spine (like other skeletal systems these days) is still building a kind of "vocabulary" for 2D skeletal systems that's different from 3D modeling and animation. They've had to because realtime 2D animation uses different techniques and has different demands to achieve convincing/good quality stuff, ie, you need to "cheat" a lot of things through reordering draw order, you need to swap images, cover up seams, redraw things, warp in specific ways to emulate a 3D volume, etc...
It's rarely good enough that you kinda-achieve skeleton animation with a 3D program using the current 3D paradigm using a bunch of flat quads or polygons, then traditionally skin them using a 3D rig. It's just debilitating.