Test suite is flaky (racy?)

Eg. https://github.com/awslabs/shuttle/actions/runs/25028936551/job/73306194391?pr=279.

Test suite is flaky. I wonder if it is due to test inference. We can look into it and try to fix it, or we can swap over to using cargo nextest (which I think we should do anyways)

Start by reproducing the failure from the linked GitHub Actions job and inspect whether test inference is involved. Compare the current test runner with cargo nextest; done means the flakiness is explained and fixed, or the suite is migrated if that is the chosen resolution.

Written by the indexing model from the issue text.