In the worst case yes, since that might change the number of words that fit in that line, and that would mean exchanging words with the line before or after that, leading to a cascading effect.
However, you might find that many of the old linebreaks can stay as they are, so I'd expect reflowing to be faster in practice.
How efficient is that under local changes? E.g., when the user (or javascript) changes one word in a long paragraph, does the algorithm require a re-analysis of the whole paragraph?