Expected Behavior of Go1 Handstand

I trained a Go1 Handstand policy using the tutorial notebook with default settings. However, I observed that the Go1 is actually jumping around the circle without standing still. I didn't see any commands were used during rollout. Is this behavior expected? What should I do if want the robot to stand still?

https://github.com/user-attachments/assets/cd72d199-8127-4736-a732-6dec1ccf531b