problem with gzip 1.5 and 1.9 on the new cluster
after some hours of debugging, we found the 1.5 and 1.9 version of gzip implies problem with the server load. huge peaks (> 1000 of load average) were observed and reproduced in the case of massive jobs using system gzip (on the new slurm cluster). It was not observed with the sge cluster, using version 1.3.
installing 1.3 gzip version on the new cluster solved the problem, but we have to use get_exec_path("gzip") in the components instead of gzip command line