nans magnier

I found an alternative method using only SOPs:

Grab a Volume node.

Name: density
Initial value: (1, 0, 0)
Go to "From Camera" and fetch your camera object.

Set both Z Near and Z Far to 1 (you can set them to any distance, but they need to be the same value to ensure the volume is flat).

Uniform Sampling: 100
Drop a Convert Volume node and lower the detail level to match the grid density you need (0.1 in my case).

Drop a Convert Line node.

Now, you have a grid that follows the camera movement and is adjustable.
It's not the best solution but can be useful in some situations (like for alembic cameras).