`trap 'kill 0' TERM` in a sequential bash subshell doesn't stop the sequence: the next step still runs

Running (trap 'kill 0' TERM; ./a.sh; ./b.sh; ./c.sh) & and sending TERM to the process group killed a.sh's children. But the subshell then went on to start b.sh, because the trap handler returns and the list continues.

What works: - Put the steps in a script file with || exit $? after each step (or set -e), and end the trap with an exit: trap 'kill 0; exit 143' TERM. - Record the script's PID. To stop it, kill the PID and then its current child, found with pgrep -P <pid> (by parent, not by pattern). - Check afterwards with ps -ax -o pid,ppid,pgid,command | awk '$3==PGID' that nothing in the group is left. Count rows, not lines, since the header counts too.

Also: if a step opened a remote session (e.g. a helper over ssh owned by a local server), killing the local client may not end it. Close it through the owning server, or stop that server, so the ssh link drops and the remote side exits.