Sucks on GPUs to have bad chips. I might have a similar problem on my Windows machine. But I have an AMD card. Just did an investigation on my own... and it looks like mine is clean! Did an overnight test... 100%. By the way, did you try some memory downclocking? Usually helps or fixes at a performance cost...
Did the guardrail introduce any performance degradation?
If not, then way better solution than using soft ECC...