I think rather than upping the timer on how often the NPCs repeat themselves, it would be better to reduce how far away voices can be heard (or at least a more dramatic reduction in the volume of the voice over distances).
I don't want to have to stop in front of an NPC for a minute or two to see if they're going to say something because I happened to arrive between comments. But that's what will happen if NPCs speak less often. Better to make it so you can't hear them until you're much closer. That will help to cut down on the crosstalk.
I'm also thinking voice volume should be linked to the character's position, not the camera's position. Being able to focus on a conversation happening down the street from you is understandable and all.. if you're Superman. Our characters are not.
