Iirc, RaveOS inspects the kernel logs looking gpu errors, then first tries to restart the miner, then the next time it reboots the system. It sounds like you want to have the miner reboot immediately? The problem is that RaveOS sends a hard kill signal, most probably terminating the miner before it has a chance to launch a watchdog script.

But, if I'm wrong and that functionality isn't enabled for you in RaveOS, you can try adding --watchdog_script=watchdog_sysrq.sh, that's our standard forced reboot script. That will only work if RaveOS bundles that with the miner in their own internal miner package. So, the bottom line is that the RaveOS support is probably better for helping you if the above doesn't work.