Formatting splits a sentence across text nodes mid-phrase, and no model
reliably reproduces a fragment's leading or trailing whitespace — a dropped
edge space glues two words together once the fragments are concatenated
back into the document. The source fragment now decides the edges.