However, that's not my takeaway from this talk. He says it was so stressful mainly because of time pressure due to fast-approaching deadlines as owners wouldn't push back the deadlines. Obviously there's a relation between starting VO earlier in development and having less deadline-related stress, but it's not clear if that compensates for the additional work due to having to re-record things and take away time from writing early in development ("the writers aren't writing any more [...] because they're recording VO and that takes all of their time.")
If he was given 3 additional months from the start to put into VO, removing the main source of stress, would he put that near the beginning, middle, or end of development? I don't think this question is answered by (at least that section of) his talk.
I think he would need to get into more detail of how process of writing no/partial audio differs from full-audio. Outside sheer scope and exhaustion, what else was rushed in order to meet the deadline? I have ideas but they would all but empty speculation. The general gist is that full VO itself wasn't a problem, but way it was handled was. That said, at the end Josh said that with proper implementation he would do it again "because it is an expectation, which kind of suck but that's what it is". Why it sucks is what interests me - my guess is that it's drawing resources away from stuff that actual makes the game click, but that's just my subjective player's perception.
