Leap smearing is awful except for rough synchronization operations that don't care if you redefine a second, and even there it isn't great even if you know the exact smearing formula because you can't easily tell if the data you're getting is pre-smeared or post-smeared, or should be smeared but isn't, and then you end double smearing or inverse smearing by accident.
Maybe it’s technologically difficult from a standards perspective, but it works well for humans. If the computer smears an extra second into my day, I won’t notice.
