Wallwork makes a living correcting texts by scientists who must publish in a language they did not grow up in, and the book is assembled out of the mistakes he saw repeat: word order, articles, tenses, punctuation, sentences that grow until the thread is lost. The examples are real and taken from technical writing, not from conversation manuals.
What I took from it
The passive hides the responsible party. "Inventory will be validated" does not say who validates, and in a functional specification that is not a matter of style: it is the exact spot where nobody ends up accountable. Since that chapter, every sentence in a scope document has a subject, in both languages, even when it comes out longer and less elegant.
The ordering rule: important first. A requirement that opens with three conditions and ends with the action has to be read twice, and a document that has to be read twice gets read once.
And the one that has saved me the most argument: a requirement with three chained conditions is not a difficult requirement, it is three requirements badly separated. Splitting them changes the conversation, because suddenly one can be approved while another is discussed.
What I left out
The entire academic frame: paper structure, reviewers, response letters. I do not write papers and do not intend to. What transfers is the standard of precision, which matters as much in a project document as in a journal, because a contract gets read looking for the gap, and the gap is usually grammatical.