I think the specifics of "defensive writing" depend strongly on the subject. In the author's subject (infosec), you probably need to defend against the "it won't work" and "the assumptions are unrealistic" schools of criticism, so you end up being overprecise and including unnecessary complexity just to preempt the critiques. In other subjects (the mathy side of TCS), you mostly need to defend against the "it's probably well-known" and "what is this good for?" attacks, so you end up providing lots of context that confuses more than it elucidates and speculating about possible applications you don't believe in yourself.
I myself am a bit skeptical of how far this "defensive writing" is to be blamed for the unreadability of papers, and even if it's such a bad thing at all when done in measure. In maths, reading papers is about swearing and asking the author just as much as it is about actually reading; yet the reasons for the unreadability are often more natural than "we need to get this past the referees" (all else being equal, referees still reward readability). New ideas rarely are born easy to explain. Ugly ducklings need their time to grow. You can wait for 10 years hoping that by then your own work has taken the proper shape to explain to undergrads, or you can write it up half-baked in half a year and let others play with it. What is better for science? The answer is far from obvious, particularly because releasing your ideas into the wild will often get them properly explained faster than trying to do it yourself.
I myself am a bit skeptical of how far this "defensive writing" is to be blamed for the unreadability of papers, and even if it's such a bad thing at all when done in measure. In maths, reading papers is about swearing and asking the author just as much as it is about actually reading; yet the reasons for the unreadability are often more natural than "we need to get this past the referees" (all else being equal, referees still reward readability). New ideas rarely are born easy to explain. Ugly ducklings need their time to grow. You can wait for 10 years hoping that by then your own work has taken the proper shape to explain to undergrads, or you can write it up half-baked in half a year and let others play with it. What is better for science? The answer is far from obvious, particularly because releasing your ideas into the wild will often get them properly explained faster than trying to do it yourself.