So, I have a best practice question. I created a few characters in their separate files. Rigged them using rigify, created a pose library for rubharb lip sync. Next, i linked that character to a separate blend file which will be my scene file, and created a proxy so as to be able to pose it. Now for initial setup, the characters are kind of different scales, so I need to scale them all to look same size with respect to the scene and also pose them initially for frame 1. So scaling, will of course be in object mode, but the transform location, should I do that in object mode or should I do it in pose mode using the root bone? Also, when using rubharb lip sync, should I generate the lip sync actions before i link the character, or should I do it on the proxy itself in the scene?
If you plan on scaling the characters at different scales for different scenes then I would Follow Jaroslav's advice. However if you just found out they were not scaled ideally and now that you see them together in a scene you want to adjust them permanently, you should go back to each characters rig file and make the adjustments there, this way you won't need to worry about factoring scale of your characters when you try to animate them in your scenes. Having to deal with scale factor of your character may complicate your animation process, or give unexpected results when running certain scripts if you use those...
You definitely want to set up the phonemes in the rig file of each character (think of them as the poses the face of your character can do always: A, E, I, O, U, BVF, smile, angry etc...). In your scene file you should only worry about animating those existing face poses to match the dialog of that scene. If you find you need to add some expressions, go back to your rig and add it there, that way it will be available for any scene you want to do in the future. Your character rig will get more powerful as you add expressions to it.