Launch near-horizontally. Have your guidance system point the nose up at 45 degrees as of say, 0.5 to 1 seconds after ignition and hold it there for the duration of the rocket burn. Then you can use as many balloons as you like!
The tractor design seems a bit iffy. Nothing wrong with a little bizarreness, but this is going to mean that you have burning hot exhaust gasses flowing over your aircraft. Not to mention the Wikipedia link on pendulum rocketry that RayG posted! Don't forget that the tractor escape system on various NASA rocketry has never been put to use, and they probably have a much bigger budget to make sure it would work properly! If you're worried about getting up to a speed where control surfaces will work, then can't you put, say, a "sacrificial" flap or two in the rocket stream? Doesn't matter too much if they burn up, so long as they last long enough to get you going in the right direction. That or make the rocket nozzles directable.
I still think balloon clusters would be an idea. You don't just have one cluster with all the balloons touching. The idea is that if an entire cluster does go bang, you still have redundant clusters. Have a heavily inflated lower set, a normally-inflated middle set and an upper set only just inflated enough that they will lift the rig on their own. You get the ballocket through all that turbulent weather nice and nippy-quick, plus footage from an upward-pointing camera of the staged balloon bursts would be pretty awesome. Plus the rig will most likely go higher, as the upper set only needs to just-about-barely carry the weight of everything. I would also think it pretty obvious that a cluster of three balloons needs to be inflated vastly less than a single balloon, and therefore will go much higher before burst.
Off the shelf rocket motors also tend to have an ejection charge in the tip, and I don't know what that would mean with them mounted in a tractor config like that. It seems like you'd need your own extra method of removing the tractor scaffold after the burn completes. Mounted to the side or underneath in a more normal fashion, you could use the ejection charges to blow your rocket modules clear in a Shuttle SRB style.
The folding wing idea is something. Perhaps have a design where the force of the air sweeps the wings back like a fighter? You're going to need something to keep wing surface area to a minimum during the rocket burn, as otherwise you will more than likely rip both wings off (or at least, limit your potential maximum speed) even up at that rarified altitude. Don't worry too much about fitting ailerons to a moving wing. Elevons work extremely well for controlling both pitch and roll, and you reduce your servo count by two. A simple V Mixer (available from most model shops for about a tenner) can turn your aileron/elevator signals into elevon signals easily enough. Just like this, and yes, that is me: http://www.youtube.com/watch?v=0x2mbEVqJW4
Finally, and I've already said this but it bears repeating: Cameras. Cameras and more cameras. On the launch rig. On the Vulture 2. Hell, on the rocket boosters too. Cameras everywhere!