We have got the same situation on two production servers (two of tens of them). Hard to tell what triggered the issue... But the server is unable to recover without a reboot.