Recommended way to randomize fixed-base articulated assets with replicate_physics=True and clone_in_fabric=True

[kankanzheli]

Hello,

I would like to ask about the recommended way to randomize the initial pose of a fixed-base articulated asset in Isaac Lab when using many cloned environments.

Suppose I have an articulated asset whose root/base must remain physically fixed during simulation. I want each cloned environment to have a different initial base position and/or yaw angle, while keeping the asset fixed after initialization.

My questions are:

When using InteractiveSceneCfg(replicate_physics=True, clone_in_fabric=True), is it possible to randomize the fixed base pose of each cloned environment independently?
If the articulated asset has a fixed joint to the world, should this randomization be done only at startup before physics replication, rather than during reset?
What is the recommended Isaac Lab approach for this case?
Using startup randomization?
Randomizing the env origins?
Creating multiple template assets and cloning each group separately?
Avoiding clone_in_fabric=True if the fixed base pose differs per environment?
If per-episode reset randomization of the fixed base pose is required, is there an officially supported method while still keeping replicate_physics=True and clone_in_fabric=True?
For a fixed-base articulated object, is writing root pose through tensor APIs during reset physically valid, or is it unsupported because the base is fixed by a constraint?

I want to understand the correct Isaac Lab design pattern for randomized fixed-base articulated objects in cloned RL environments.

Thank you.