What's the problem this feature will solve?

I do use --last-failed to rerun flaky tests. This works perfectly fine in formal testing of libraries, however when I use the same machinery for jupyter notebook-based tutorials I need to rerun all previous cells, too as tests are not standalone.

(I use the nbval plugin to collect the cells as tests)

My other option would be to use pytest-rerunfailes, but that plugin reruns the failures straight away, and most of my cases are flaky as they query a remote server. So the approach of --last-failed would be preferred, aka to have a second test tun that only runs the problematic ones.

Describe the solution you'd like

Have an option to be used along with --last-failed that would collect all the tests from the files contained a failed test

Alternative Solutions

Additional context