Why would you…

…set text documents to open in an office-suite word processor by default, rather than a text editor?
…set no-word-wrap as default in your text editor in the era of wide screens?
…do away with the status bar when it tells the user whether the window is fully displayed vs merely larger than the screen?
…put a search field in the main page of your web browser when all you’re going to do when people try to type into it is relocate the cursor to the address bar?
…design window buttons [Minimize/Maximize/Close] to not respect Fitts’s Law?
…automatically hide scrollbars in the era of wide screens? especially when it never occurred to you when the normal screen was the size of a postage stamp?
…restore sessions by default? Especially in a web browser?
…design a resizable dialog to resize the section that won’t benefit from resizing, rather than the section that would?
…set a button to produce the click output “Configure the application first” rather than the configuration dialog?

BONUS! Hey, Linus Torvalds! Why would you… introduce an error message, viz, “the capability attribute has been deprecated”, such that when you g**gle it (since it means nothing by itself!) the closest thing to a hit that’s returned is the actual kernel code that produces it:

static ssize_t disk_capability_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
dev_warn_once(dev, "the capability attribute has been deprecated.\n");
return sprintf(buf, "0\n");
}

…which contains not even a comment? If the message is meaningful only to kernel devs, and you’d have to be one to understand this, why not hide it behind a kernel command line debug switch?