available here. It includes a good explaination of the roll problem on the prior flight and how it was quickly resolved.

The unwanted roll on the Sept. 29 flight was essentially a high-Mach snap roll to the right. Like many high-speed craft, SpaceShipOne loses directional stability at high Mach numbers, and this effect is greater at low or negative angle of attack (AOA). The departure started when the angle of attack became slightly negative at Mach 2.7 and 120 KEAS, which is at roughly 120,000 ft. Directional stability dropped and the craft yawed about 8 deg. Strong dihedral effect coupled the yaw into roll, and the nose pitched up about 15 deg. as well, starting a snap roll motion that was uncomfortable for Melvill.

The motion quickly settled into a pure roll that was more tolerable, with Melvill's head near the roll axis. The rate was about 180 deg./sec., and Melvill countered the roll with left rudder and rudder trim. With the spaceship climbing nearly vertically at roughly 200,000 fpm. the equivalent airspeed was dropping rapidly, and he was able to reduce roll to about 140 deg./sec. by the time aerodynamic controls became ineffective.

While Melvill found the pure roll physiologically acceptable in itself, he was concerned about it coupling into a disorienting tumbling motion, and he also wanted to bring the ship back in pristine condition for the required second X Prize flight. Ground control suggested he shut the motor off before the full 89-sec. burn, and after the altitude predictor indicated the ship would make 100 km. he closed the oxidizer valve at 77 sec.

At the standard 5 KEAS Melvill raised the feather, then started using the cold gas thrusters to zero the roll rate before apogee. Rutan noted the flight demonstrated the thrusters could stop a 140-deg./sec. roll, but there was not much gas left and "we don't need to try that again." Melvill did a roll during the glide back to the field.

The pitch profile was changed on the Oct. 4 flight to avoid negative AOA. Normally engineers want a strong early pullup to quickly point the craft near vertical, but maintaining positive AOA will eventually put it on its back. The Oct. 4 flight had a less aggressive pullup to avoid this problem while keeping AOA positive. With a long motor burn of 83.9 sec. it was able to top the prior flight.

"We address anomalies like NASA but with world-class promptness," Rutan said. "The last five days have run on 12-hr. cycles--sort it out, run it on the simulator, lots of meetings. We had no mechanical squawks from the last flight and only had the roll problem to deal with."

World-class promptness indeed.