Github actions spontaneously failing

[jack-fs]

CI runs like https://github.com/gradientinstitute/causal-inspection/actions/runs/4932765788 have started failing, where the same commit passed at https://github.com/gradientinstitute/causal-inspection/actions/runs/4626515640.

It appears to be related to module name resolution: in particular, we see errors like:
ModuleNotFoundError: No module named 'simulations'

Further digging is required to find the root cause

[AlistaiReid]

In the CI logs, it looks like the environment / install is timing out - I'm seeing the last line of the log being "Error: The operation was canceled."

It could be the base images are so old its a major update to set up the envirnment

[jack-fs]

The CI has two jobs: one for python 3.8 and one for 3.9. The 3.9 one is failing because of a ModuleNotFoundError and I think if one fails then the other is killed

[jack-fs]

I have just managed to replicate this on my machine by creating a completely new venv rather than just forcing reinstalls in the one I was previously using.

It required an upgrade of pip: 22.3.1 -> 23.1.2. I am suspicious that this is related to the move to pyproject.toml, which I believed started to be enforced at pip 23