Also keep in mind that it’s not just RAM that can have issues, but also the slots it sits in.

I once had a server-grade workstation (dual-socket AMD Athlon MP, 8 slots with 4Gb ECC REG apiece). Each piece of RAM tested perfectly OK by itself in the default primary slot, but failed consistently in one secondary slot and intermittently in another. The slot hardware (pins) were fine, but something elsewhere in the motherboard had broke.

Which is why you also test each slot with a known good stick of RAM, just to be absolutely sure.

Instead of losing 8GB of precious RAM, you could map out the bad memory with a Linux GRUB boot parameter that tells Linux not to use that part of the memory.

I've successfully done this with a Dell Precision machine that we needed as a build machine. Twelve months later on, still works and we do seriously stress it with builds.

https://www.gnu.org/software/grub/manual/grub/html_node/badr...

BTW, I absolutely love Pepper and Carrot. Keep up the great work.