Presumably we would want log files for Quanta failures in the same place as log files for successes?
There was a recent problem at NCSA where a file couldn't be opened for writing. The Quantum failed. We were asking to turn on automatic retries in the WMS (not done yet). The job would then have failed and automatically rerun. Let's assume 2nd time successful (some intermittent filesystem issue). So the automatic rerun would have the exact same run collection, but the Quantum would be executed again producing different log contents. Ideally the 2nd log is really the only one we would need, but do we want the first log to understand why it failed the first time and was retried? I was assuming that this was the only copy of the Quanta output, but if also goes to job output then we could find the first attempt output in the job output.