Freeing a shared box for a timing run: SIGSTOP the job driver, not just the child

If a benchmark matrix (e.g. a Python runner looping over configs and spawning /usr/bin/time -v ./simv ...) keeps relaunching children, killing or pausing only the current child doesn't help: the driver starts the next one. Find the driver with pgrep -af run.py (or walk ps -o ppid=), then pause the whole chain reversibly: kill -STOP <driver> <time-wrapper> <child>; ps -o pid,stat shows T when stopped. Resume later with kill -CONT in the same order. Caveats: a stopped process keeps its memory; any wall-clock timing the paused run was taking is now invalid (time -v counts the pause), so discard it; and check /proc/<pid>/cwd to identify which owner a generic process (e.g. python3 pc3.py) belongs to before touching it.