Emergency shutdown:
Yes, I can add something like this.
However what exactly should happen after the temp-limit is reached on one of the GPUs? should zm terminate immediately?
Also don't forget, a user-space program can't guarantee you to react properly with respect to this (the hardware or the driver can get unresponsive) - however it should work in most cases.
I'd say stop using overheated GPU and keep informing about it in console, like Claymore's or EWBF's behavior. I know, the driver or hardware will shutdown it in case of overheat, but 100° is not healthy.
Maybe --temp-target can handle sudden rise of temperature and we don't need the emergency shutdown at all, but it's just not obvious from the description.