A recent change no longer async throws these errors, but relying on that to fail the test was always brittle.

It would be great if we could add a device to expect error logs and fail tests if they don't do so.