Off the top of my head, I can't think of anything that will help you right
away. One idea might be to use different SLOTx_USER (formerly VMx_USER), so
if you found out that a job was running on slot1 you could figure out which
directory matched that job.
Another idea is to add a unique identifier to each job's ClassAd that you
could query on, and then transfer a file with the same name to the execute
side. Then, if you want to find a job, just look for the directory that has
the uniquely-named file.
You could also stream your output files back, or use a wrapper script and
have Chirp send them back as the job progresses.
-Erik