Question about Unitree Go2 model

Hey, I hope you're doing good!

I'm trying to replicate the go1 walking task of mujoco playground with the go2 model present in this repository and I noticed that it has a different impratio (being 100 in the go2 and 1 in the go1) even tho in the mujoco documentation is recommended to keep it low. I tried to maintain that number but the problem is that it brings numerical inestability into the situation for bigger timesteps, is there any reason for this bigger impratio in reality? Is there a way to have the same effect without breaking numerical stability?

This is the model i'm using (added some go1 sensors that were used in the go1 walking task, change the integrator to keep a numerical stability and increased the iterations with respect to https://github.com/google-deepmind/mujoco_playground/blob/main/mujoco_playground/_src/locomotion/go1/xmls/go1_mjx.xml)

```

```