Since 1972, timekeepers have periodically added a leap second to Coordinated Universal Time so that atomic clocks and the actual rotation of the planet do not drift too far apart. The practice was built around a basic assumption: the Earth’s rotation is slowing, so any correction will involve adding a second. That assumption no longer holds. In recent years, the rotation has accelerated slightly, and the next correction could instead require subtracting one second. That development has alarmed software engineers, distributed-systems architects and IT operations teams, because no major global computing platform has ever been asked to process a negative leap second.

Key facts

• The CGPM will vote in October on a plan to make UTC continuous from May 20, 2027.
• Avoiding leap seconds would let UT1, or Earth-based time, drift away from UTC by as much as one hour.
• A negative leap second has never happened before.
• Analysts estimate that the probability of a negative leap second could reach 30% by 2035.
• Past leap seconds have caused outages at Reddit, LinkedIn, Qantas and, more recently, Telstra.

A turning point for global timekeeping

The proposal to be considered by the General Conference on Weights and Measures (CGPM) in October is designed to avoid a dangerous scenario. Rather than prepare the world’s software for a correction that has never been attempted, the CGPM wants to make UTC continuous from May 20, 2027. In practice, that means leap seconds would no longer be inserted or removed. UT1, the time scale based on the Earth’s rotation, would be allowed to drift from UTC until the difference reaches one hour. The action would mark a major departure from the system in use since atomic time was linked to astronomical time.

The idea has been under discussion for years. The CGPM warned that a negative leap second has not previously been applied and that it poses a high risk of causing anomalies and disruption to critical infrastructures that are largely unprepared. The cost of preparing industries that rely on time synchronization, the body said, would likely be similar to the investment required for the millennium bug, or Y2K. At a 2025 workshop, experts studying the Earth’s rotation concluded that the probability of a negative leap second would increase rapidly, reaching 30% by 2035. That projection helped turn a hypothetical concern into a vote of practical importance.

Jeremy Roberts, senior director at Info-Tech Research Group, said authorities never expected to reverse time. “We have been adding leap seconds for decades, basically to keep atomic time in line with observed time, but this is the first time we’d have to take one away,” Roberts said. “Computer systems aren’t designed to work this way.” He argued that the current proposal lets the two clocks go out of sync instead of forcing continuous adjustments, which would make behavior more predictable for companies building and operating infrastructure.

Roberts also acknowledged that the plan is not perfect. Allowing the difference between UTC and UT1 to grow as large as one hour could create trouble for future generations. But he said indecision could cause fragmentation as different entities move to different time standards. In his view, fragmentation may bring a separate set of problems for global commerce, networking and software development.

Why a negative leap second is a hard problem

Most enterprise software has never been tested for a negative leap second. When a leap second was positive, system clocks would either repeat a second or jump forward. With a negative leap second, clocks would need to skip a second, which means a system clock would appear to move backward. That is more difficult for software to handle. Many applications rely on monotonic clocks and strictly increasing timestamps. A backward jump can break optimistic locking, security token generation, scheduling algorithms, logging sequences and audit trails.

Frank Dickson, principal analyst at Dickson Research, said the timekeeping change is bigger than most technical decisions. “This is the biggest change to civil timekeeping since the leap second itself was adopted in 1972,” Dickson said. He noted that the Earth’s rotation has historically been slowing, so the leap second system only ever had to handle one direction. Now that the rotation is speeding up, a negative leap second would require subtracting a second. “Nobody has ever run that in production, at global scale, on the systems the world actually depends on,” he added.

There is also a history of ordinary leap seconds causing real-world failures. In 2012, a single positive leap second took down Reddit, LinkedIn and Qantas’s booking system. It also triggered a race condition in the Linux kernel that spiked CPU load across servers worldwide. The failures occurred because the situation was rare and few teams had planned for it. More recently, a timekeeping-related problem disrupted the Telstra network in Australia. Those incidents are often cited by

Source: Network World News