This test uses a parallel shell, and (in this passthrough) sometimes expects a future read to be causally consistent with the other shell's update. This is not a valid assumption.

While we could re-write the test in a way that would allow the main thread to wait and observe the parallel shell's write, it doesn't seem worth it to modify the test and gain that coverage. If such coverage of parallel shells and causal consistency is desired, it should probably be written as it's own standalone test.

As part of this, we should do a quick scan to see if there are any other tests that might expect causal consistency with a parallel shell.