The GPU temp is likely caused by a failing cpu fan, unplugged or missing external brainbox fan, dirty filter grates, debris,etc. On newer OS versions you can correct the issue causing the overheat and reboot it and the error clears even though it is listed as a "non recoverable error" On some OS versions though, once it triggers the overheat and you reboot or key it with the audit key it will cause the safe storage event queue full which is an actual non recoverable error and will need to be reoptioned.
The event queue is not very big (mb) so once errors stack up (flakey door switch, constant thermal reporting etc) it kicks it to a Hard out of service to force you to resolve the issue.