[Bullet] Running project gives recover from penetration error.

[ca3games]

Godot version

3.3.2 (stable official)

System information

Kubuntu 21.04 (Hirsute) GLES 3 renderer.

Issue description

After compiling a project without issues and compiling without error, after 2 minutes on average I get this error:

E 0:00:05.563 recover_from_penetration: Condition "shape_idx < 0 || shape_idx >= cs->getNumChildShapes()" is true. Returned: false
<Fuente C++> modules/bullet/space_bullet.cpp:1294 @ recover_from_penetration()
WALK.gd:41 @ Physics()
FSM.gd:29 @ _physics_process()

The WALK.gd:41 line is:
FSM.Root.move_and_collide(Vector3(dir.x, 0, dir.y), false)

Not sure what causes it, but the project runs without error around 1-2 minutes just fine and then spit that error.

Steps to reproduce

It happens after 1-2 minutes of playing the game.

Minimal reproduction project

Flor Game.tar.gz

[pwab]

I can confirm the error on my system (Windows 10, Ryzen 5, RX 570). But I already get it when pushing one arrow-key once (so not time depended). After receiving the error I just waited a few seconds and got another one which then floods the console log:

E 0:00:12.303   get_quat: Basis must be normalized in order to be casted to a Quaternion. Use get_rotation_quat() or call orthonormalized() instead.
  <C++ Error>   Condition "!is_rotation()" is true. Returned: Quat()
  <C++ Source>  core/math/basis.cpp:791 @ get_quat()
  <Stack Trace> FSM.gd:25 @ _process()

This error seems to be time depended because I wasn't able to directly reproduce it.
The errors are pointing at

func Physics(delta):  # <- called on _physics_process
	var dir = FSM.direction * FSM.vel * delta
	FSM.Root.move_and_collide(Vector3(dir.x, 0, dir.y), false)  # <- Error 1

and

func _process(delta):
	current.Update(delta)
	var target_point = Root.translation + Vector3(-direction.x, 0, -direction.y)
	var old_basis = Root.transform.basis
	Root.look_at(target_point, Vector3.UP)
	var target_basis = Root.transform.basis
	Root.transform.basis = old_basis.slerp(target_basis, delta * 10.0)  # <- Error 2

[rburing]

Fixed by #59864.

[fabriceci]

I tried with the mrp and the error is gone as @rburing said.