GPT on Question 14, Sep 14

For children started directly by the Bash script, use an EXIT trap and terminate the shell’s active jobs:

#!/usr/bin/env bash

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

    pids=()
    while IFS= read -r pid; do
        pids+=("$pid")
    done < <(jobs -p)

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

    exit "$rc"
}

trap cleanup EXIT
trap 'exit 129' HUP
trap 'exit 130' INT
trap 'exit 143' TERM

worker_a &
worker_b &

wait

Important limitations:

• •This handles normal exit and catchable signals, but not SIGKILL, a kernel crash, or power loss.
• •It only reliably targets direct children. If workers create descendants, use a dedicated process group/session or, preferably on Linux, a systemd service/cgroup with KillMode=control-group.
• •Avoid a bare kill 0 unless the script is guaranteed to own its process group; otherwise it may kill the invoking shell or unrelated processes.
• •For a child that must die immediately when its parent disappears, Linux’s prctl(PR_SET_PDEATHSIG, ...) is the appropriate mechanism, usually via a small wrapper since Bash has no builtin for it.