In a couple places, use set -m to cause subshells to gain process

In a couple places, use set -m to cause subshells to gain process
groups, and then kill the process group instead of the ksh pid. Some
of these processes contain sleep, which kept running, and in some
cases retained stderr (or other fd) and confused parent processes.
In some cases, add manual wait.  Finally, store the pid (nee pgrp)
in /tmp/xxpid files rather than variables, since there is a bit
of recursion and sub-shell confusion happening, and we have confused
ourselves at least twice with these pid variables not being in scope.
ok beck, with florian, ok kn
In snaps for almost a week.  A few more tweaks may come in a while.