Another thing I'd like to add, since the topic of power keeps coming up. I am currently using an external case (https://www.amazon.com/Kingwin-Enclosure-Internal-Backplane-Optimized/dp/B01BMJ1WD6/) to house my HDDs (the WD and ST), whereas the cache drive (SSD) is connected to the onboard sata port.
I'm using an intel NUC7i3BNH - specs here: https://ark.intel.com/content/www/us/en/ark/products/95066/intel-nuc-kit-nuc7i3bnh.html
To power these drives, I'm using an AC -> molex adaptor (12V/5V) and then I use a sata power splitter because the kingwin enclosure needs two sata 15pin power.
Does any of the above sound suspicious that could lead to this? To eliminate the kingwin enclosure as the source of the problem, I connected the drives directly to the JMB card and then powered the drives using my sata power splitter directly. I still got the same issue, hence I ruled out the external enclosure to be the source of the problem. However, power is something I haven't switched. I did use the same way to power another 3.5inch disk and connect to my computer via USB (to copy some data - I have a sata to USB cable), and it worked fine. So I didn't think the power supply can be a problem. However, I didn't power 2 drives together.
Just trying to think through what could be the root cause. If the JMB chip is known to work with unraid, then it could be my motherboard (nuc) + this card? Which is weird, because I did read success stories of people using the same card (iocrest) with NUC and FreeNAS/Windows. Which brings me to the point of what it means when we say "it works". Can I ignore the errors if here are no read/write errors while building the parity and data copying? Or am I limited in speed to the link resest and I cannot rely on the setup?