Somewhere in my travels, I saw a LilyPond source file in which the author had created a voice that contained only formatting directives and no notes, so that the other voices would contain only notes and no formatting directives. This "technical" voice contained all the staff markup, barlines, dynamics, tempo changes, etc. I remember seeing that and thinking it was a good idea.
Now I'd like to implement it but I can't remember how it was accomplished or where I saw it. Can someone point me in the right direction? Is this a recommended practice?