We have performed a series of highly-instrumented experiments examining corner-turning of detonation. A TATB booster is inset 15 mm into LX-17 (92.5% TATB, 7.5% kel-F) so that the detonation must turn a right angle around an air well. An optical pin located at the edge of the TATB gives the start time of the corner-turn. The breakout time on the side and back edges is measured with streak cameras. Three high-resolution X-ray images were taken on each experiment to examine the details of the detonation. We have concluded that the detonation cannot turn the corner and subsequently fails, but the shock wave continues to propagate in the unreacted explosive, leaving behind a dead zone. The detonation front farther out from the corner slowly turns and eventually reaches the air well edge 180degrees from its original direction. The dead zone is stable and persists 7.7 mus after the corner-turn, although it has drifted into the original air well area. Our regular reactive flow computer models sometimes show temporary failure but they recover quickly and are unable to model the dead zones. We present a failure model that cuts off the reaction rate below certain detonation velocities and reproduces the qualitative features of the corner-turning failure.