collada support is great if animation is supported, but probably is not enough for complex animation, like mesh deformations etc (water, clothed characters etc) plus it needs a lot of work to support the supported features which is not obvious if exporters supports them.
for v2 or i hope earlier there is no need for you to implement an animation system
a lot of effort is needed ( hierarchies, keyframe interpolations, bones with weighting etc etc... )
You could set specifications for your animation import format (just a simple ascii one) and let the community work.
You/We could start incrementally - starting from a general scaling, translation, rotation node which can be attached to objects and cameras - then introduce an attachment to this node with camera FOV, depth of field parameters etc..
these nodes or probably your existing nodes could get text file inputs for per frame variables tweaking. ( one row of variables per frame for example)
for deformable objects the exporter could save just the vertex positions x,y,z of the deforming object per frame.
or specify a more holistic format, instead of importing motion per node, the user could be loading to octane the static scene through obj and then load a second file with per frame motions of objects,lights,camera, materials, deformations etc.
by loading a second animation file it would be more convenient since the user will be able to tweak textures lighting etc from octane and achieve much more accurate results than through the materials specified by the obj mtl format.Plus there will be no more problem with internet connection failures.
the possible pipeline:Open octane->load scene->tweak scene->load motions->render anim
animations is must, octane is fast
but loading and saving objs to hard disks is slow and it is too pity loosing all that speed gained from octane to slow disk I/Os