Autotuned but crashed

https://review.px4.io/plot_app?log=3a1b9c62-149a-46cf-8b3c-13aaaba2bd60
This link is a record of a successful flight lasting 22 minutes and 15 seconds. To check stability in various flight postures, while flying by rotating fast the roll and pitch stick in a circular motion at 8 minutes and 35 seconds, large vibrations occurred in the quadcopter and it almost fell, but it barely regained its balance on its own.

https://review.px4.io/plot_app?log=88e1694d-1f62-47ef-8167-b6577e3dd1f5
This link is a log of the second flight, about 40 minutes later and finally crashed. During this flight, while performing the same maneuver as before, keeping the nose to the east and moving quickly along a circle with a diameter of about 1 m, the multicopter began to experience large vibrations, forcing me to lower the throttle of the remote controller and bring the flight altitude close to the ground. It was lowered, but it continued to vibrate and failed to land safely.
It had previously crashed during a similar type of test flight. At that time, I thought the cause was the use of the default PID value.
What I’m curious about is whether the vibration and fall that occurs when this maneuver is performed is an inherent limitation that the PX4 cannot overcome.
It was a calm and warm day.

I’m just pinging @bresch here.

@bigbear117 It’s a bit strange because during most of the flight, the tracking looks quite good and it gets suddenly bad.

Is the minimum PWM value high enough? (i.e.: are the motors spinning when arming the drone in stabilized mode with 0 throttle?) The current and vibration look really low on arming.

I confirmed that when the aircraft was tilted more than 60 degrees in an idling state with the propeller removed, the motor on the raised side maintained idling and the motor on the lowered side spinned to its maximum.

There are gaps in the quadcopter’s folding arm joints and landing skids. This morning, I measured on the ground the vertical displacement of the booms where the motors are mounted and found that all four booms were bobbing up and down by about 4mm. My assumption is that any maneuver that causes the motor to swing below free play will result in loss of control. What would your opinion be? The link below shows vibration analysis.
https://review.px4.io/plot_app?log=19769389-049f-4584-bccc-9156bfe1775a