I wrote a collada -> tvm in about a week with the examples out there, unfortunately its for a specific flavor (sic) of collada (3d pdf) so probably not much use. also only does static objects and not animations etc.
Native collada support would be great, but it wouldn't be an easy task as everybody does things a little differently.
If somebody wants a collada solution i'd be willing to have a crack at it, but would need an incentive