> It's probably a timing thing,
> where we kill off the one instance but need to insert a do nothing loop
>(or a few of them) to wait long enough before we try to start it up
> again. The problem is probably more noticeable with the faster machines...

There is a Rexx SysSleep function as well as several compiled executables that call the DosSleep control program function.  Those would probably work better than a "do nothing loop" if you want the cpu cycles free for other uses