The way I would re-construct a vocal line, would be to match (from a database of natural sounds) samples close to sounds found in the English language; i.e. the "ch" in "bach" could be enhanced with the breaking of bark (a crass example, mind you would have to be very subtle in the task). Hell, if one wanted to go all out, there could be two separate Sylvari dialects, one for the "plant" Sylvari (a mild speech), and one for the "bark" Sylvari (more aggressive tone).
Of course not every syllable would be enhanced (or could they with vocals already recorded; would, to a good degree already dictate what syllables could even be enhanced, if at all in some cases) or it would probably be inaudible; but a subtle application would really make their speech stand out from the rest imo; less is more. It would depend on how what is being said, to determine whether the enhancement should be on the attack, decay, sustain, or release. I think I would clearly understand, "Where is bach?", if an enhancement slightly lifted the "Wh(e)re" and accented the "ba(ch)".
Enhancing hard consonants would surely give them a "tribal" feel (I forget the name of an African tribe (if they are even African) who uses "clicks" and other sounds in speech). For the Sylvari, wind going through a hollow reed, for example could enhance open sounds (i.e., "who"), etc.
Though we come down to practicality; the vocals more than not have already been mostly recorded and mastered; adding such sounds would take a lot more work (taking a step back to original files); then you have the big problem if you wanted to provide the vocals in other languages (it could still work, but you have to reconstruct the natural sounds to best fit each dialect; which would take forever); and all this with the many movie hours they have of Sylvari script.
A program could maybe be written to automatically add stuff, but it would have to work damn good (it would probably sound robotic...); the subtlety of a human crafting each natural sounds with its English language counterpart imo would be the best approach in pulling the effect off; yeah probably be the only way to go, because then you take into account the emotional tone of the voice, etc.
Localization should seriously stay on a purely text based level, otherwise it'd take way too many ressources away from actual progress with the game.
No matter how you all feel about everybody should understand and speak proper English, there is no way ArenaNet/NCSoft will compensate for losing money in GW2 sales, when they don't supply localized versions.
Martin Kerstein answered a question about German voice actors in an interview with Rush (IIRC) a few weeks back and somebody talked about the Spanish version on the ComicCon panel. So they are in.