When a node fails, schedule a future ping to check again.