Sometimes, after disks failures we can see in the kernel logs that errors returned by osd_trans_stop() are actually from an earlier processed request because dr_error is not reset.