Flaky tests: cargo test --workspace produces inconsistent failures

[foxfromworld]

When running cargo test --workspace locally on uutils/coreutils, I observed that the set of failing tests changes between runs. This suggests that some tests are flaky or environment-dependent.

Environment

• OS: Ubuntu 22.04
• Rust version: rustc 1.93.0
• uutils/coreutils commit: 6a942ba

Observed failures

Different runs produce different failures, for example:

Run 1:

• test_df::test_df_arguments_override_themselves
• test_df::test_df_compatible_sync
• test_df::test_df_conflicts_overriding
• test_df::test_df_masked_proc_fallback

Run 2:

• test_df::test_df_masked_proc_fallback
• test_tail::test_follow_truncate_fast
• test_touch::test_touch_changes_time_of_file_in_stdout

Run 3:

• test_df::test_df_masked_proc_fallback

Run 4:

• test_df::test_df_masked_proc_fallback
• test_touch::test_touch_changes_time_of_file_in_stdout

Run 5:

• test_df::test_df_masked_proc_fallback
• test_tr::test_truncate_applies_before_complement_with_class

Expected behavior

Tests should be deterministic and either consistently pass or consistently fail.

Actual behavior

The set of failing tests varies between runs, even without code changes. This makes it difficult to validate local changes before submitting a PR.

Question

If I submit a PR and the CI/CD pipeline fails due to these flaky tests, what is the recommended way to proceed?
Should contributors re-run CI, mark these tests as known flaky, or simply note that failures are unrelated to the PR changes?