Hi Tilman, welcome and thanks.
Well, the PID controller works on Euler angles, which are vulnerable to effect. Quatos works on quaternions, which don't suffer from that issue. It's not that you can't (or shouldn't) switch from rate mode to PID angle mode ever, but if the craft attitude is just right (or wrong), it may lead to the gimbal lock problem while it tries to recover past ~90* tilt. I can't tell you how likely this is to happen (or not).
Since AQ already tracks its attitude using quaternions, we just need to get rid of the angle-based PID controller.
I actually started playing with a simple quat-based PID controller, but it was a bit too simplistic and I ran out of time for now.
Absolutely the best way to test all this is using the simulator. Trying stuff that would be stupid to do in real-life is great fun (and educational). The main situations where I can force gimbal lock is when one of the quad X arms is pointed sharply down (so, near 90* tilt and roll) and I initiate yaw -- at some point in the spin it'll go into gimbal lock and start rapidly rotating around one of the pitch/roll axes (and then drill into the ground arm-first).
Cheers,
-Max