Conversation

For most of human history, the "single source of truth" for what time it is has been the Earth's rotation. We didn't even have a clock that was more accurate than the Earth until 1921, the Shortt-Synchronome clock.

When UT was first created, it was based entirely on the rotation of the Earth. Which meant the UT second was variable, because the Earth doesn't rotate at a constant rate. This wasn't good enough for astronomers, who needed a uniform time scale. So they came up with Ephemeris Time, based on the average day length in 1900, which in turn was based on the tables in Newcomb's Tables of the Sun from 1895. The way they got from tables of numbers to practical values like the count of Cesium oscillations is really interesting (hint: it involved the moon). If you want to know more, look up "ephmeris time" on Wikipedia.

Even after the invention of atomic clocks, the Earth was still considered the single source of truth for time: time broadcasts were kept in sync with UT (Universal Time, then the same as GMT) by stepping them 20ms at a time, but their rate was still based on the Ephemeris Second. But 20ms steps only really worked if systems could tolerare one second being shorter or longer by a little bit, and people didn't really know in advance. With more and more systems becoming dependent on synchronized time, this became impractical, so UTC (Coordinated Universal Time, with the letters arranged so they're equally wrong in English and French) was born. UTC is based on TAI (International Atomic Time, actually correct in French), but it's offset by an integral number of seconds, currently 37 (it started at 10 seconds off since TAI was defined to be contiguous with the older Ephemeris Time and that's how far TAI was off from UT, rounded to the nearest second).

Until now, UTC has been kept within 0.9 seconds of UT (really UT1 as UT varies depending on the current orientation of the Earth's axis and where you are) by adding or subtracting one second at the end of June 30 or December 31, just before midnight UTC. The last one was added at the end of 2016, so we had 2016-12-31 23:59:60 UTC. Leap seconds have only ever been added, though it's also possible to have a negative leap second. It's just never happened.

As you can see, that last leap second was almost 8 years ago. Earth's day has stayed very close to an average of 86400.0 seconds since then, in part due to climate change melting glaciers and causing the planet's rotation to speed up. If we continue with leap seconds, it seems likely that the next one would be the first negative one. Very few systems have been tested with a negative leap second.

But there aren't really good reasons to keep UTC in sync with UT1 anymore. While it's convenient for astronomers, the relative number of applications where 0.9 seconds is "close enough" has dwindled. It'll definitely require astronomers to make some changes, but astronomy has gone from the most important application of UTC to one of a great many applications, most of which are actually harmed by having to deal with leap seconds.

So it's very likely that the 2016 leap second will have been our last. Not really because of the changing rotation of the Earth, but because they were already thinking about getting rid of them in 2016, and it's just time.