This ensures that its resource usage metrics get recorded somewhere rather than getting lost. (It also fixes an error when called from a nested logging context which completes before the bg process)