Oh ok sorry, it seems like I didn't get the point of this thread quite well..
I thought it was about optimal settings that can get you best results in less time..
As for animation I think you should first try and find the optimal settings for different shots in each of your scenes.
Animation is nothing more than sequences of images, a renderer as in this case Corona has no effect on the way a video is encoded it just gives you the images.
There are many programs you can use to composite these images and make the final video.
As for the settings I believe they're pretty straightforward if you don't use the "Debug UI".
If you want to get an unbiased solution which yields the best results you could try leaving everything to default (PT+PT) and then changing only parameters that might be needed for your scene.
For example if you have a complex refractive object you would have to raise "max ray depth" parameter in order to
change the maximum number of times a ray can be refracted and get the correct result.
As for "path tracing samples" 16/32 (few times 64) is best, in order to provide an even distribution of sampling..
And so on..
Now I mentioned "Bucket Method" because it gives you adaptation which I believe is what you would look for when you want to have fast results with even noise distribution across your images. It's closer to the way Arnold Renderer works which is used in films lately, hence used for animation.
Still excuse me if I fail to understand the context of your question.
Cheers..