The argument is from the generator implementation point of view. Using regular escape is much simpler. Unlike regular escape, you still cannot use CDATA for attribute values. CDATA might quickly become a footgun because it can give you a false sense of security. Regular escaping is much more universal and also works with HTML content / attribute values.

> I'm not sure how CDATA makes "debugging confusing," and in any case using CDATA blocks inside an article seems like a pretty rare case; like I said, I haven't done that myself.

Debugging can be confusing when you actually encounter that closing sequence, the text becomes much less readable and seems kind of broken. With regular escape the content is more distinguishable (even though harder to read) from the structural XML. Actually, its rarity can be more of a problem because you might never know about it and not even handle it in your own serializer at all. The "magic" of CDATA is dangerous. You might not believe it, but many developers still don't do any proper escaping when injecting text in DOM elements. They often do element.innerHTML = "Some untrusted text". I have seen such things countless of times.