The last-lap failure that exposed a Williams software flaw

The issues that arose at the start of the Bahrain GP in Malaysia last weekend put an uncomfortable spotlight on Formula 1’s 2026 power units.

At Sepang, a software bug shut down cars on the formation lap as a perfect storm of circumstances that hadn’t been anticipated tripped up the system. It was another demonstration of just how complex the current power units are.

However, as is so often the case in Formula 1, it was a situation that had an uncannily similar precedent, albeit at the end rather than the start of a race – one that occurred in much simpler times, in the very early days of computers running racing cars. And even 35 years on, Nigel Mansell’s late demise in the 1991 Canadian GP stands as one of the most infamous last-lap retirements for a race leader in the sport’s history.

That season signalled the start of a golden era for Williams after three disappointing seasons since the departure of Honda. Adrian Newey had joined from Leyton House the previous summer, and he designed a brilliant aero package for the FW14, while Patrick Head led the mechanical side as the team introduced a paddle shift for the first time.

In the background, the R&D team was pursuing a range of high-tech projects, notably the active ride test programme that was led by software wizard Paddy Lowe.

Meanwhile, having announced his retirement at the 1990 British GP, Mansell was persuaded to return to the Williams fold after a two-year absence, with clear number one status above Riccardo Patrese.

From the off, the new car was a competitive package. However, in the early races it was hampered by a gearbox issue that was eventually traced to a shaft expanding at high temperatures. With that addressed, Mansell logged his first finish of the year with second place in round four in Monaco.

At the next race in Montreal, things really clicked. Patrese earned the team’s first pole of the season, while from second on the grid Mansell took the lead and totally dominated the race to the extent that he was able to back off in the closing stages.

Then as he rounded the hairpin for the final time he slowed and acknowledged the crowd, and the car promptly rolled to a halt. A delighted Nelson Piquet snuck past to win for Benetton, while Mansell was far enough ahead of most of the field to be classified sixth.

In the aftermath, Mansell took a lot of flak for the premature celebrations, and it was widely suggested that he’d somehow switched the engine off, adding to his distrust of some elements of the media.

“They said I switched it off and all that kind of crap,” he recalls. “The ‘box just went into neutral and would not come out of neutral, and basically from what I saw in the cockpit, there were some slow shifts a number of laps before, but for whatever reason it lost power, either electrical power or the solenoid just stopped working.

“I should have kept it in fifth gear and gone around the hairpin in fifth and burned the clutch out or something, and got home. And yeah, it’s a painful one to lose when you’re flying, and then have the press say stupid things to you.”

So what really happened that day? The reality was more complex than it may have appeared at the time, as various team members explain.

“He was miles ahead, and on the last lap, he was slowing down and waving to the crowd,” Head recalls. “We had a rocking lever on the paddle shift, and pulling on the right was up, and pulling on the left was down. And because it was a sort of rocking lever, if he’d pushed the lever with his fingers on the right, he would have changed down.

“But he was waving to the crowd with his left hand, coming down to the hairpin, and the engine revs got very low, like 2500rpm or something, as he slowed down. And when he pulled for a downshift, the engine revs were so low that the alternator didn’t keep the battery up at the right level, and the engine just cut out and died.

“He was not to know that the engine wouldn’t keep running if the revs got so low, because basically we had a tiny little battery and it wasn’t enough to keep the engine running.”

Future F1 technical director Geoff Willis had followed Newey from Leyton House to Williams, mainly in a pioneering CFD role.

He confirms that Mansell was caught out by circumstances that had not been anticipated.

“I think he did a rather lazy downshift, which was refused by the box,” says Willis. “The revs then were a bit on the low side. He tried again, and I think we ran out of hydraulic pressure, because it had an accumulator.

“At low revs you won’t have enough pressure. And I think from that point it spiralled down, because then the engine revs are dropping, the car’s slowing, the pump pressure’s dropping. You don’t have enough pressure, and I think he just didn’t have time, or he didn’t realise what was happening fast enough, to pull the clutch in.

“And of course the clutch was hydraulic, so on the same system, and that was the final killer. Had he dipped the clutch earlier he could have then got the engine revs back up, he would have got the hydraulic power; he could have shifted down the box.

“It was a system failure in the sense that we didn’t properly protect against that.”

Mansell was criticised for going slowly, but it’s worth recalling that he’d had costly gearbox failures in the first two races. While there had been a fix, it was perhaps understandable that he took a cautious approach in the closing laps in an attempt to protect the gearbox, especially after a glitch started to chime in.

“He had done everything right,” recalls his race engineer David Brown. “And he was beginning to experience a little bit of a downshifting issue, in that it wouldn’t always downshift the first time, just because it was dog ring damage, and it happens.

“But because he was cruising, it was happening a bit more, and in the software, there was something which said that effectively the blip for the downshift and the ability to make the downshift was all based on the RPM at the point when you pulled for the downshift.

“He was cruising around so much and changing gear relatively late, because he’s trying to get the gearbox to get through the race, and trying to be kind to it, that at one point it didn’t go into gear several times. And by the time it was ready to go into gear from a sort of software point of view, it was going so slowly that the RPM at which it would have blipped would have been lower than the idle speed of the car.

“At which point it switched it off as a safeguard, which was built into it, and it was something that never happened before. But the problem was that the software had a hole in it, if you like, which permitted it to switch the engine off if you were going very slowly and when you made a downshift.

“And that’s what it did. It just switched the bloody thing off. And it took like 15 seconds of Paddy writing a bit of code for it not to do that again…”