Oho, so Galli-Mariť was third choice for the role, and Marie Roze thought the character
was scabrous, not the opera. Well, there's still some of that going around; what role is it Natalie Dessay won't sing because the character isn't nice? But I'm glad to see Bizet got his "six additional first-upper-voices and four second-upper-voices" for the women's chorus.
Still, the interesting point in McClary's book is that the score was sold for publication two months before the premiere. Those two months were the period of intense revision during rehearsal, including the 13 versions of the "Habanera". So does that mean the "Habanera" in the score is not the one Galli-Mariť sang? It's more likely Bizet re-submitted the score to the publishers once the changes were nailed down. But that doesn't explain the reappearance of the high passages.
Andrew, thanks for finding McClary.