I would love to have 100% VO. I think that it makes a big difference in helping the player suspend disbelief and become fully immersed in the narrative. Games are an audio/ visual medium, which require the player to take direct control of an avatar in a simulated world where all of the sensory data is pre-produced by the developer. This arrangement functions best when the game world's basic mechanics replicate our expectations- we see and hear as we would in the real world. If you have a 3D world with sound effects etc which comply with this approach, but render dialogue as text, the result is a jarring subversion of expectations. This is why I don't think the book analogy is particularly helpful, since that medium expects almost all sensory experience to be generated internally by the reader.
I also feel there is a risk in using old fashioned game mechanics, even when they worked well in games we loved. The problem is that we become used to new standards as they develop in any product or service. For me, this happens to the extent that I now prefer not to go back and play my old games, because I know that my changed expectations mean that they won't live up to my memory of the original experience.
Of course, I think the counter argument has been made very well in this thread. I hope the devs can find a good position on the topic relative to cost and flexibility.