A day is a day, Earth rotation defines it. What we can do is decide how long a second is and tweak it to make a day 100k seconds long or 86.4k ones.
It's not, hence leap second corrections to our current earth based imprecise observed solar time based on mean solar days (which are not apparent solar days).
Of course even if it were regular there's that pesky difference in rotation relative to what now??
Sidereal rotation time isn't equal to solar rotation time (mean or apparent).
Time .. less straightforward than most imagine.
I think it's more trying to make things which vary fit in a "you shall not vary" square box that is the problem. Technically, this metric/decimal method makes more sense for what we're trying to do, but it's less of a "time" thing than it is a "let's have the same numbers every day so we can agree on when synchronous events need to happen." To _measure_ elapsed time, using a fixed unit such as a second is perfectly fine.