I'm a bit confused on where the characters are starting from in a time stream vs. time distortion sense. Let me see if I understand this before proceeding to mechanics.
There will end up being three time streams: the original, the distorted, and the devestated. Using the hypercurved 5-d space allows them to step outside of the time stream they're currently in, and somehow distinguish that there were other timelines that were different from the one they're in now.
So at the beginning of the game, the engineer reveals the original timeline and some minor details about the distorted timeline, which is the timeline the characters should be in now.
So the guy who Knows stuff is able to look down the current timeline and see where it diverges from the original timeline. The Engineer moves them down the timeline to roughly that point, at which time the guy who Connects things puts the Speaker, along with any Made items, into the timestream where he can then interact with, and hopefully fix, stuff. The whole time the Seer is taking notes about everything that happens.
Let's assume the Seer is taking notes because everyone will forget most of what happened when they get back, hence the single 3x5 notecard to remember stuff. Except for the Seer who can write as much as he wants.
So we've established that each character has some sort of action that is uniquely his, and that allows him to futz with things. The Connector could put the Speaker or the Maker's items into the wrong spot in the timeline. The Speaker, being more or less on his own, has any number of opportunities to hose things up. The Maker can provide incorrect or inoperable items to the Speaker. The Knower could lie about where they should go or the outcomes of actions. And the Seer could alter any aspects of the mission logs so that when everyone's heads are spinning after the trip back the Seer can say "Oh, don't you remember. You're the one that dropped the modern dollar in the Speaker's pocket." Of course, the Seer's forgotten all sorts of stuff too, so whatever he wrote becomes canon.
The characters supposedly fix the timeline and return to their own time to discover that something has gone horribly wrong. They are now forced to renavigate back down this new third timeline to discover who the traitor is. The traitor's job is not to get caught and maybe throw the blame at someone else.
This is all assuming that the characters still somehow know that the timeline they're in is not the timeline they started from. That's where I'm unclear. How do they know the timeline is different if they didn't start out in a hypercurve space when the timeline changed in the first place? I'm guessing for story reasons that this all works out somehow. Sorry, I haven't read the original thread for this game.
I'm also unclear on how the characters, rather than the players, take actions without the other characters knowing what they did. How does the POD work? Is everyone standing next to each other? Is it like a spaceship with different rooms? Can the Connector switch out an item without anyone noticing? Can the Speaker do something without the Knower knowing? Can the Maker "accidentally" break something without it being obvious? Can the Knower tell the Engineer to take them to a point in time that's incorrect? I'm going to guess that for story reasons these sort of actions are possible, because otherwise the Traitor would have a really hard time screwing things up with everyone watching.
Assuming all of this is correct, and that the Traitor can take actions without being noticed, let's move to mechanics. (And I must say, this is a really cool concept so far.)
In order for this to work, I agree with the importance of knowing when the traitor is chosen. He needs to be able to set things up from the beginning, so I would imagine that's the time to pick him out. This could be done using 3x5 cards passed out from the GM. Each card has the player's Function on it, but one of them also says Traitor, and of course this is kept secret.
Players also need a way of taking actions without everyone else knowing what they're doing. This could also be accomplished by cards. Each player could have a handful of cards with their Function written on them, and to do their function that player hands in a card and states, or maybe writes on the card, what he wants to do. The Traitor would have all the different Functions, of course.
Just as easy would be sidebarring with the GM. Each time a character is going to use a Function he steps away with the GM and tells him what he's doing. The GM comes back and provides the results. The characters then huddle and decide what to do next. Once they decide the next course of action there can be another round of sidebarring/card passing/whatever.
The mechanic has to somehow image the characters' abilities to take actions without notice so that they can be sneaky and the Traitor can actually hose people without them knowing it, and having used their own Function against them make it look like they did it. Again, this could be cards, sidebarring, note passing, etc. This would work really well on an online chat system with PMs going between the players.
I know this takes away from the collaborative story creation element of the game, but telling a mutual story where one person is trying to screw his buddies takes a lot of effort on the part of everyone else not to metagame and somehow rationalize a way that they knew who the Traitor was.
Anyway, that's my spew on the topic. I believe you have a really solid idea here if you can find a way to manage the activity.