Hi!
Okay, as for the render times - we experimented a little bit. The last shot in the sequence, with the gliders and the explosion in the hill, was rendered straight in TG2 with Motion Blur (0.3) in 1280x720. Render times were about 6 hours per frame.
We installed the maximum numbers of render licenses allowed with on TG2 Deep Animation, so the sequence took a few days on the render nodes and my own machine. I started two instances on my machine to use two processors for each running TG2-instance. For the future I´d wish unlimited render licenses (because the TG2 render times are just pure horror, so the drawback could be minimized a little bit for renderfarm-owners) and the availability to hook it up to real render distribution systems, like Autodesks Backburner for example.
For the shot with the car on the marsian surface we tested another approach. We rendered the sequence WITHOUT motion blur (2 hours max per frame), in the end rendered a detailed Micro LWO export from a high camera and loaded this geometry into 3dsmax. After a little rework on all the existing gaps this surface could be used to get a camera WSM projection from the main camera with the rendered TG2 sequence, the texture set to "mirro" instead of "tiling". So the edges, where motion blur appeared without a existing TG2 image, could be "faked". And then everything was rendered with Vray Motion Blur - saved a lot of time, but does not work with all landscapes, rather with planar ones like this. And of course not with all camera moves.
These were our two approaches and for upcoming shots now we know which one to choose to get the result as fast as possible.
cheers,
Steffen