For anyone reading this thread: This is a known issue. PID oscillation is the cause. The oscillation can be so fast and thus visibly so small that it is difficult to see. PID oscillation can come from a couple different places.
The reason is that LP code has a feature where if the throttle is low and it needs a _lot_ of motor thrust for stabilization, it will effectively increase the throttle, just so it can have more stabilization to try to keep the aircraft level. It will also reduce the throttle if you have the throttle stick at 100% for the same reason. It's easy to see that no stabilization is possible if the motors are at 0% or 100%.
Previously, if the motors were at 10% and it needed +-20% for stabilization, it would increase one to 30 and decrease the other to -10. Well the motor can't run -10. It runs zero so you get e.g. some yaw when all you want is roll.
Now throw in some reason for an out of control PID oscillation and no matter where the throttle is set, it is driven to 50%, just to be able to apply maximum differential (stabilization) control.
The solutions are:
- balanced props (don't just skip this, it is the most important)
- increased GyroTau (GCS -> Configuration -> Stabilization -> Expert -> GyroNoiseFiltering)
- reduced PIDs (experiment to find out which, suspect the D terms)