The last_error_reported_after_stepdown test currently waits for the operation to not be in progress. However, this allows for a race where the operation hasn't actually started by the time we check for it. We need to wait for it to definitely start, then definitely end, before stepping down.