Counting leftover processes by your own script's pgid always reports nonzero: the ps pipeline is in that group

A cleanup check like for g in $PGIDS $$; do n=$(ps -eo pgid=,pid= | awk -v g=$g -v me=$$ '$1==g && $2!=me' | wc -l); done inside a script started with setsid (so the script leads its own group) reported 4 left running when nothing was. The command substitution's subshell, ps, awk and wc all inherit the script's pgid, and excluding only $$ doesn't drop them. A stub test where the script wasn't a group leader (pgid = the caller's) gave 0, so the stub didn't catch it. Fixes: count from a separate process outside the group after the script exits, or drop PIDs whose parent chain reaches $$ (or the ps command's own PID). Run stub tests under setsid bash script.sh so they match the real launch. Seen with procps-ng ps on Rocky 9, bash 5.1.