totally free from bugs is a much higher and impossible standard to meet than the bare basics of "don't solder a SSD onto a motherboard that you know will wear out in a few years".

It's more fundamental, to continue the software analogy, it's more "how did they possibly miss that" error like shipping a device that's supposed to have mariadb listening only on localhost, but it actually listens with no authentication on all live network interfaces.

You’re picking a single mistake as somehow unusual. It’s easy for the original spec to have been fine, some change happens that nobody considers in that context and boom you get an issue. Depending on usage pattern a SSD can be expected to last a 20 years, but changing the usage pattern by say storing more data isn’t going to raise a red flag saying you now have an issue.

Someone turned on excessive Linux logging and boom an issue shows up early on some a Tesla Models. Chances are it wasn’t even intended to enter production that way.