Re the peppering of ‘A’s and ‘THE’s that R’s puzzles get, I guess the main difficulty is that adding them in just to plump surfaces seems slightly unfair: when they’re a part of the wordplay (as in ‘the compiler’ = THE/ ME) it’s a different matter of course.

BTW, Sil, have you remembered FT14057 yet?

The indefinite article’s potential for ambiguity is exploited by lots of setters! I suppose it can sometimes be used wrongly but I don’t see any sign of that here and 19a isn’t Rufus just being arbitrary or perverse. “False friend” is a rather weak, boring anagram indicator – it works but false doesn’t really mean mixed up or redone, whereas “a false friend” seems to me to imply “a wrong version of ‘friend’” so it conveys the anagram idea better. That’s what I thought when I was solving it anyway and it’s why I singled it out on the other site as a particularly good clue. The construction is unusual and potentally confusing (i.e. cryptic), but effective. But the main point is that, while the surface is better with the article, the cryptic instruction is meaningful whether you have it or not, just as you could always clue “De Niro” as either “actor” or “an actor”. Both are correct and the setter tries to use the most confusing or interesting version. A good setter, like a good writer, should surely exploit alternatives and uncertainties, in which English is unusually and famously rich, not avoid them. And the solver should be aware of them – s/he should know for instance that we can say both “I’ve been talking to miserable Toby” and “I’ve been talking to a miserable Toby” and that the difference between the two is quite subtle (it’s like in the clue – “a miserable Toby” carries a stronger implication that he wasn’t always thus). This doesn’t mean you have to like the clue, but I certainly think it works.