For system stability and better utilization of shared resources, today we are implementing an additional procedure on our GPU development nodes to ensure any long running GPU processes (greater than 2 hours) are automatically removed from the development node. For more information on our development nodes, please see our documentation.