GPT on Question 14, Sep 18

In Bash, the usual approach is to record child PIDs and clean them up from an EXIT trap:

#!/usr/bin/env bash

children=()

cleanup() {
    status=$?
    trap - EXIT INT TERM

    if ((${#children[@]})); then
        kill -TERM "${children[@]}" 2>/dev/null || true
        wait "${children[@]}" 2>/dev/null || true
    fi

    exit "$status"
}

trap cleanup EXIT
trap 'exit 130' INT
trap 'exit 143' TERM

some_command &
children+=("$!")

another_command &
children+=("$!")

wait

Important limitations:

• •Traps do not run if the script receives SIGKILL, crashes, or the machine loses power.
• •Killing recorded PIDs only kills immediate children, not necessarily their descendants.
• •Avoid kill 0 unless the script owns a separate process group; otherwise it may kill its caller too.

For an entire process tree, start children in separate sessions/process groups and terminate the groups:

setsid some_command &
children+=("$!")             # session leader / process-group ID

# During cleanup:
kill -TERM -- "-${children[0]}"

If children must reliably die even when the Bash parent is killed with SIGKILL, Bash alone cannot guarantee it. On Linux, use a small launcher that sets prctl(PR_SET_PDEATHSIG, SIGTERM), or run the workload under a supervisor such as systemd, which can kill the whole control group.