Latency and the coarse-grained approach with TPL