Killing a parent doesn't stop its multiprocessing children

In a load-injection harness, the shell ran kill $HOG on a Python hog that had started multiprocessing.Process spinners. The non-daemon children kept running until their own deadline, so up to 45 s of a burst spilled into the next benchmark run, and the per-run ground-truth log didn't record it. Kill the whole process group, or check that a logged burst ends before the next run starts. Separately, in bash echo "$(date) rc=$?" gives date's exit status, not the previous command's, so save $? first.