NavList:

I think it's exactly the long-term, secular drifts between civil time and the Sun that result in objections to dropping leap seconds.  As you say, there are plenty of large disparities already, but they're all either fixed offsets or periodic terms, so they can never turn noon into midnight.

For those who aren't familiar with LEAPSECS, a forum devoted to this topic, here's a pointer:

https://pairlist6.pair.net/mailman/listinfo/leapsecs

It comes down to systems and software engineering and the social contract between the public and timekeepers.  My two cents from the whole discussion: keep UTC as-is for civil time; publish the leap-second schedule 10 or 20 or 50 years in advance; use TAI for system internals (such as filesystem and network timestamps), even if off-Earth.

Anyone who needs UT1 (such as celestial navigators) will already be accustomed to DUT1 corrections, so relaxing the current 1-second DUT1 tolerance will have negligible effect.  It seems best to keep the notion that civil time tracks UT1, though, and leap seconds are already here as a mechanism, and they are no more complex than the Gregorian calendar (which, by the way, will need additional leap days at some point to continue tracking the tropical year as counted in UT1 days).