The motherboard apparently mine for 20-24 hours, then freeze mining and ssh is un reachable (connection reset by peer).
The only way to control again the system is to power cycle
Temperatures of the gpus are quite high and they start and stop to cool down, could this be the reason?
probably yes, there is a temperature limit to each miner with the default values when not configured, the miner has watchdog that reset or restarts the app in order to avoid GPU damage. same is true to OS, I bet you are using a unix/linux base system as you do SSH.