Yes, that would be doable, I've done it before for stills. It should only take another pass and then a batch operation in PS to apply it but the DOF in this previous version was TG-rendered.
To get around the issue of the outer edges of each render bucket blurring, I had to make each render bucket the size of the final render, limit TG to work on just one core per render and use new instances of TG to do different sequences of the animation simultaneously in this configuration. Waiting on a 1280x720p version at those settings wasn't practical at all, especially with the hang(pc freeze)time between frames with the high AA settings in the micropolygon renderer that I needed to use.
I might try and do the long-way-round Z-depth pass before I process the frames into a video, I might not...
