Hi Rodolfo,
Thanks for reporting the issue.
This almost certainly an out of memory error, but it is possible there is an issue with the machine, or the specific node your job landed on.
A couple of things:
1. It sounds like it is reproducible, which makes a specific node error seem less likely.
2. Have you tried increasing the available memory? There are two more memory options. They make the run more costly, but may allow it to get through.
3. Is there really no difference between these runs and the ones that work in terms of expected memory footprint?
4. If you are stumped, you can send me the _JOBINFO.TXT file for a job that worked, and one that did not. I will investigate.
Mark