The one mistake with UNIX is that UTC is used. It should have been TAI, and leap second adjustment should have been handled by the timezone package (but at the time people weren't worried about leap seconds, so it's understandable). As it is, the system clock has to change. This is as bad as MSDOS time (and other systems which put localtime into the actual RTC). Using UTC as the time reference, independent of local time, was nearly genius. With TAI it would have been pure genious. No, wait, using an epoch time of the Big Bang, and step size Planck time, now that would have been pure genious. It doesn't take as many bits as you may think..
Civil timekeeping is based on UTC because we want clocks and calendars to remain synchronised with the sun. It's not just astronomers who care about whether the sun is above the horizon. If you think about it, the solar day is the one unit of time that almost everybody cares about and pays attention to. If you screw up the day, you also screw up the week and anything else that is based on counting days. Why would you want to do that?
But offset of UTC for TAI is like 37 seconds or something poxy like that. Nobody cares about this, the sun moves too slowly for that to make any difference.
There's been talk of setting UTC=TAI for quite a long time and I think it'd make sense. It'd take so long for TAI to drift from the actual rising and setting of the sun that we'd probably all have standardised on Swatch Beats by then anyway.
Yes, the difference is 37 seconds now, but it's growing quadratically. It's quite possible that we'll change the way we express times and dates, but it's almost certain that we'll want them to stay in phase with the sun, and we'll probably want to stick with the SI second and with something like TAI. My guess is we'll continue to want a simple relationship between TAI (or its successor) and the civil clock time, something like the current relationship according to which they are a whole number of seconds apart. That implies that we'll need something like leap seconds (or leap minutes perhaps). There doesn't seem to be a reasonable alternative.
In 1950 astronomers pointed out that there would have to be two kinds of time, one to agree with calendar days and one to be as uniform as possible. Arguments over subsequent decades inexplicably decided that there could only be one kind of time specified by international agreements, and we ended up with a choice of two out of three characteristics in what we now call UTC. https://www.ucolick.org/~sla/leapsecs/picktwo.html
I feel like throwing in a leap hour (basically tz shift) once in a millenia would be more reasonable solution, if for nothing else than letting future generations deal with it instead of trying to futilely pre-empt problems that are not really problems yet.