Start with the size of the swap. A leap hour fires only once atomic time and rotation time have drifted 3,600 seconds apart [5]. Every leap second applied since 1972 sums to 27 [6]. The replacement therefore triggers at roughly 133 times the total correction the current mechanism has ever made [17], and at the historical average rate, 27 seconds across the 44 years to 2016, the first one would land somewhere near 5,900 years out [18]. Macdonald's estimate of hundreds or thousands of years is, if anything, conservative [15].
That average will not hold. Over millions of years tidal friction from the Moon slows the spin, but on short scales it wanders, and since about 2015 the planet has been turning faster, a change attributed to currents in the liquid core [11]. Every one of the 27 adjustments so far has added a second [6], while the one now in prospect would remove one [4], and that asymmetry is where the operational risk sits [20]. The code paths that swallow a 61-second minute have all been through at least one live event, whether they smear the second across a day, step it at a specific minute, or stop trading through it [9]. A minute 59 seconds long has never had to be handled by anything, anywhere [4].
The 30% figure deserves the qualification Nature's report gives it: it comes from scientists who study Earth's rotation, without a named model or a stated interval [12]. As a forecast that is soft. As a statement of tail risk it is sufficient, because the complement is only a 70% chance that nothing needs to happen [19], and the budget for handling an event that has never occurred gets spent in advance either way. The International Bureau of Weights and Measures, which published the resolution, says a negative leap second could disrupt critical infrastructures and would be expensive to mitigate against [13], and Macdonald reports apprehension about which equipment might fail [14].
What the leap hour does not explain is the near-term payoff, which belongs to abolition by itself. Once no leap second is coming, the gap between one firm's gradual smear and another's instant step stops being a compatibility surface [9], and the discontinuity Macdonald objects to, in a system intended to be continuous, stops arriving at all [10].
On this evidence, abolition is the right call and waiting until 2035 buys nothing [1][2]. A subtraction that has never been performed cannot be rehearsed [4], and the replacement event cannot reach any machine now racked [15]. The condition attached to that view: somebody eventually has to specify what crossing 3,600 seconds does to a clock, and the account of the resolution published in Nature gives the threshold without the mechanism [5]. Macdonald grants that a leap hour may still present technical difficulties, and rests the reassurance on financial systems working differently thousands of years from now [16].
