I keep getting hit by this bug:

• The version number: 20131222
• The bugid: loadavg_file-r: /home/scdata/.parallel/tmp/loadavg-61104-mach5

The loadavg file doesn't exist.  The pid and host vary, so it's not just one isolated instance or a broken host---although our network is known to be flaky.  Unfortunately I am not at liberty to include the input files or programs involved and I have not been able to reproduce this by simpler means.  The invocation of parallel looks something like this:

parallel --gnu --pipe -L 1 -n 1 --retries 3 -S $sshlogins --joblog joblog --load 85% -- $program

The hypothesis I offer is that occasionally we have a network glitch and parallel fails to retrieve a remote host's load average; perhaps this condition is not expected?