My observation was initially rejected with the "unclear" mention.

I will try to be clearer in explaining my context:

I deliver an application through Docker to ensure cross-platform portability.

My development environment is Mac with an Intel processor.

The application code is identical regardless of the platform.

What can change from one platform to another is the hardware, the operating system, and the environment that runs the Docker containers.

Originally, the stack included in my Docker image was as follows: a) Centos 8 b) Python 3.9 c) Pandas 2.2.2 d) Numpy 1.26.4

When running this image on my Intel Mac, there were no execution problems.

When porting this image to a first Linux system with podman 1.6.4 (instead of docker), still no execution problems.

However, when porting this image to a second Linux system with podman 4.0.2, I encounter this error "TypeError: Cannot convert numpy.ndarray to numpy.ndarray".

I then completely modernized the image stack:

a) switched to Red Hat 8 b) adopted Python 3.12 c) adopted Pandas 2.2.3 d) Numpy 1.26.4

The problem is still present on this last platform (absent on Mac)