I agree with you about the preeminence of the text in establishing an interpretation, but, as far as I'm concerned, text-setting by a composer is usually very much tied up with the sound of the words, particularly the way vowels work in different registers in a given language, not to mention the stress-patterns of the language and how these convey meaning often to a comparable extent to the semantic content. ... Then there's the question of how you translate the text, given that it's possible to force all the aforementioned features into a different language with different prosody, vowels and consonants: should it be done according to the (for example) kind of English spoken at the time the opera was composed? or at the time it's being performed? or some other time to accord with whatever the production is trying to do?
I was going to jump in here with my translator's hat on, but Richard's done such a good job of it that I don't think I need bother, except to add that, from what I understand of French/English opera translation much of the problem lies in the different stress rhythms - one's DA-dee-DA-dee-DA-dee and the other is dee-DA-dee-DA-dee-DA - so I can quite understand Wainwright being unable to convert the work into English just like that. For my money, I think translating an opera libretto faithfully and in a way in which it fits perfectly into the music must be about the hardest translation work going, and is why I tend to favour having them sung (properly, of course
) in the original language with surtitles.