That's right. We already have the logic to render per-word.
If justify is set for halign, we can do the first pass as usual (left or
right, doesn't matter), then justify in the second pass (we'll know
exactly the size of each word + width of the line.)
Mathieu