GP brought the point up that addresses this: "why deprive readers of extra clarity offered by this formatting?"

So if this is something that's valuable when reading or editing material, why not extend that to the final, rendered output?

To me, this smells of micro-optimization that's not well thought through: where are the boundaries between being able to edit vs being able to read? If we make every word be on a line by itself, you can use remove-line command in your editor, and diffs will automatically become word-diffs, and it would encourage writers to limit sentences to clearer sentences by fitting them into one ~50 line/word screen. By using double newlines, you can still keep "semantic" newlines too. Wouldn't that be appealing? "No" is what I would say.