[BUG]: Hang in `test_select_drop_nulls_aggs[ray-aggs1]` in cudf-polars tests

**Describe the bug**

https://github.com/NVIDIA/cudf/actions/runs/32816649081/job/97706230710 contains a very large traceback from a test failure.

Using the timing inserted in the logs, the last successful test was at `2026-08-25T06:35:14.7069533Z`. Our timeout fired at `2026-08-25T07:58:12.1429258Z`, 80 minutes later. The python traceback indicates that we're inside `test_select_drop_nulls_aggs[ray-aggs1]`. That looks like a hang, rather than just slow nodes that we've seen previously.

Start with the linked GitHub Actions job and the `test_select_drop_nulls_aggs[ray-aggs1]` test, using the reported timestamps and Python traceback to reproduce or localize the hang. Done means the test no longer hangs until the timeout and the relevant cudf-polars test run completes successfully.

Written by the indexing model from the issue text.