[Bug Report] ImportError: libopenblas64_p-r0-0cf96a72.3.23.dev.so` when running any Isaac Lab script that imports Replicator

[Sukeysun]

Environment:

• OS: Ubuntu 22.04
• GPU: NVIDIA RTX 5060
• NVIDIA Driver: 580.126.09
• Python: 3.11 (conda environment: env_isaaclab)
• Isaac Lab: Latest version (installed from source)
• Isaac Sim: 5.1.0

Description:
Running any Isaac Lab script that uses Replicator fails with an ImportError indicating a missing OpenBLAS library, even though the library file exists in the environment:

ImportError: libopenblas64_p-r0-0cf96a72.3.23.dev.so: cannot open shared object file: No such file or directory
ImportError: Error importing numpy: you should not try to import numpy from
        its source directory; please exit the numpy source tree, and relaunch
        your python interpreter from there.

The library is actually present at:

/home/lenovo/miniconda3/envs/env_isaaclab/lib/python3.11/site-packages/numpy.libs/libopenblas64_p-r0-0cf96a72.3.23.dev.so

This occurs with official examples:

python scripts/reinforcement_learning/skrl/train.py --task=Isaac-Ant-v0

Root Cause:
The library exists but RPATH/RUNPATH in the NumPy extension modules may not include the .libs directory, or there's a conflict with other OpenBLAS installations in the system path.

System State:

• Multiple OpenBLAS versions present in conda environment and system paths
• The specific .libs directory containing the required library is not being searched at runtime

Request:
How can I ensure that NumPy's bundled OpenBLAS library in numpy.libs/ is correctly discovered at runtime? Is there an environment variable or configuration needed for Isaac Sim 5.1 to properly resolve these dependencies?