I had a similar debacle, when I managed to corrupt a btrfs file system to point it wouldn’t mount again…

I was preparing it to have as my main system on bare hardware. I had accidentally mounted the same block device simultaneously in the host and guest: kablamo silent corruption and all 5 hours of progress lost.^* :(

*shred the guest VM, host was ok.

On arch, UEFI boot vars are mounted at /sys/firmware/efi/efivars. It’s unwise to rm -rf them…

Until you would have to replace a HDD: +23 hours of nerve racking RAID repair time for 10TB drive at 120MB/s Even with some advanced (like ZFS etc.) system you can’t go around the fact the HDDs are slow.

And when the HDD fails, you can’t read it. It’s toast. Some cheap non-volatile memory devices are like this too, but good ones go into read-only mode and you can at least attempt data recovery from them if no better option is left.

I’m liking that it is possible get cheap+good 1TB NVMe devices for less than 100€. The consumer SATA market for large SSDs (capacity over 1TB) is unfortunately quite dry. I need replacement for HDDs and even if the speed is capped by SATA bus it would be an massive improvement.