Following the power surge incident, compute nodes (may) have scratch access problems. We are currently checking them, and rebooting when possible/necessary.