RDRAND is not the problem. The design failure is in assuming collisions don't exist. That's a poor design decision made by the developer that is directly the

RDRAND is not the problem. The design failure is in assuming collisions don't exist. That's a poor design decision made by the developer that is directly the responsibility of the developer. If this was a one-off bug in systemd, it could get a pass, but it isn't a one-off. Try to boot a Linux system using systemd when something is just slightly off, like one block device that's normally available is missing or when networking doesn't come up. systemd does not handle failures well, and it doesn't handle them because the main developer makes no effort to test failure modes. This incident is yet another example of this pattern of development style. When I write code, I have to make an effort to test what happens during unexpected failures. I expect the developer of the most critical component to booting a modern Linux system to do the same.