Once the ICU units in Manaus's hospitals were fully occupied, the crisis spilled out into the corridors. Patients lying sideways on makeshift beds and breathing from oxygen cylinders lined the hallways as friends and family watched over them.
Outside, others were less fortunate. Some died at the hospital doors after having been denied entry. Some who decided to stay at home, where there was a bed available, choked to death. The city's oxygen supplies were exhausted.
The chaos that swept Manaus in January — doctors performing manual ventilation on critical patients, families rushing to buy oxygen and graveyards struggling to keep up with burials — sounds like the result of a tragedy unforeseeable and unpreventable.
Yet experts had warned of the danger of rising COVID-19 cases for months. And this was the second time it enveloped the Amazonian city, not the first.
Thought by many to be the first city in the world to overcome the pandemic, Manaus has become a cautionary tale of what can happen when politics and science fail, leaving the novel coronavirus to spread almost unmitigated.
"We tried to warn people of the scale of the problem but they thought we were crazy," says Jesem Orellana, a local epidemiologist who pleaded for a lockdown to avoid the disaster.
"But eventually it was too late for anyone to deny. When the hospitals became battlefields they didn't need convincing anymore."
A crisis foretold
Manaus is an industrial city and the biggest in the Amazon rainforest. As well as a bustling home for 2.2 million people, it is a hub for many indigenous communities connected by river.
When the coronavirus arrived in March 2020, it spread rapidly. Crowded housing, a lack of clean water and a large informal economy made the state of Amazonas the epicentre of Brazil's epidemic.
Bulldozers carved out trenches for mass graves and dusty graveyards were dotted with seemingly endless lines of blue crosses, each one marking a life lost to COVID-19.
Such was the size of the outbreak that many epidemiologists, politicians and the public deemed a second wave of cases impossible.
A study published in September 2020 by researchers at the universities of Oxford and São Paulo estimated that 76 per cent of the population had been infected with the coronavirus. So few had avoided contagion that the remaining population could not sustain another large outbreak.
In becoming the worst-hit city in Brazil —where over 235,000 COVID-19 deaths have been recorded — Manaus was thought to have become the first in the world to reach herd immunity.
In April and May 2020, coronavirus cases and related deaths dropped rapidly. Throughout the summer the epidemiological curve flatlined.
"After May and June people were going about their lives normally in restaurants and malls," says Paulo Lotufo, an epidemiologist at the University of São Paulo.
"The media and the governor were very, very excited to say that Manaus was a pandemic-free zone."
But in August, the flatline in cases stopped and hospital admissions began creeping up.
Calls go unheeded
In September, Orellana and other experts got an audience with local decision makers to try and halt the virus's spread before it spiralled out of control.
The then-mayor, Arthur Virgílio Neto, listened. He closed the city's principal beach, extended the opening hours of COVID-19 hospitals, and encouraged people to work from home. He requested state police support to shut down the city for two weeks.
But the request was not well-received. It was "absurd", Brazil's President Jair Bolsonaro told the media. The policy of lockdowns was "over".
The governor for Amazonas state, Wilson Lima, rejected the proposal out of hand. The following day, primary school classes resumed.
It was then that the yoyo of tightening and loosening restrictions began, relieving and panicking local experts forewarning of a second crisis. It was also when they believe the battle against COVID-19 was ultimately lost.
"If Manaus had implemented the lockdown in September you probably wouldn't know me, we wouldn't be discussing the new variant, and even more likely we wouldn't be living through this awful disaster," Orellana says.
Studies show that the reopening of schools spread the disease, and they also "triggered a sense of security in people", says Lucas Ferrante at the National Institute for Amazon Research.
In October, Ferrante and others published a call for action in a paper titled 'Brazil's policies condemn Amazonia to a second wave of COVID-19'. As was to be the pattern, it went unheeded.
Bolsonaro vs Science
Since Bolsonaro was elected in 2018, Brazilian scientists say they have been under attack. Bolsonaro has publicly criticised scientists and cut their institutions' funding.
The right-wing firebrand has downplayed the threat of COVID-19, describing it as a "little flu", and encouraged the use of unproven cures. He has propagated conspiracies that there was no crisis in hospitals and discouraged Brazilians from being vaccinated. "If you turn into a crocodile, it's your problem," he said.
In Manaus, like São Paulo and other cities where local governors and scientists have proposed lockdowns, Bolsonaro has radicalised his supporters against them through outbursts on social media and at rallies.
"Manaus is symptomatic of the fact that the federal government is not leading the [pandemic response] effort," says Anya Prusa at the Wilson Center's Brazil institute.
As in the United States, how people vote often predicts whether they wear a mask or practice social distancing, Prusa says. In Manaus, which voted Bolsonaro in 2018, the use of masks and social distancing has been consistently lacking.
Then, images of the Oxford study suddenly started appearing in local meetings and the press.
The study examined antibodies in samples of 1,000 blood donors in the city and found them in three-quarters of the population. The conclusion was in the title: COVID-19 herd immunity in the Brazilian Amazon.
Orellana and Ferrante doubted its credibility, but like previous warnings they went unheard. The study, Orellana says, backed politicians who did not want to implement more stringent measures in Manaus and set the tone for reckless behaviour encouraged by Bolsonaro supporters.
By mid-January, hospital staff were experiencing a cruel form of déjà vu as hospitals collapsed once again under the strain of COVID-19 admissions.
"You want to save a person's life but then you see that there is no oxygen and if that person dies there are 50 more waiting in line for an ICU," recalls Renata Santos, a nurse working in one of Manaus' COVID-19 intensive care wards.
"I will carry that memory with me for the rest of my life."
A bleak case study
Manaus's second tragedy painfully disproved any theory that it had reached herd immunity.
Rather than the first city to be liberated from the restrictions of the pandemic, Manaus is a bleak case study of the suffering that must be endured to pursue herd immunity through infection — a goal which many epidemiologists deem impossible.
But why the study's conclusion was so wrong is less straightforward.
Lotufo, Ferrante and Orellana believe the prevalence of antibodies found in the study was simply too high. As blood donors are more likely to pick up infectious diseases than the general population due to their behavioural profile, they skewed the results higher.
Others are more confident in the study's findings, however.
Deaths in Manaus are roughly what one would expect with the study's estimated infection rate, says William Hanage, a Harvard epidemiologist. Studies in Peru and Colombia's Amazon found similarly high results, too.
The second outbreak could be caused by antibodies declining more quickly than expected after infection, meaning second infections fuelled the second wave. But the simultaneous emergence of a new variant in Manaus is "suspicious", he says.
Like the so-called South Africa and UK variants, P.1 has multiple mutations of concern on its spike protein — the part the virus uses to latch on to human cells and which antibodies target to neutralise the virus. Studies have found that one of those mutations, E484K, also found in the South Africa variant, likely allows the virus to evade existing antibodies.
"[But] it's hard to disentangle the effects of waning antibodies from a variant that is better able to sidestep immunity from prior infection," Hanage says.
"Both could be happening, and it is reasonable to think that anyone with waning immunity will be yet more vulnerable to reinfection with such a variant."
Too late to prevent the spread
Brazil's health minister, who is under investigation for negligence in the Manaus crisis, has said P.1 is three times more transmissible, but studies have not been published to back the claims.
Tests are currently underway at Sao Paolo's Butantan Institute to establish the efficacy of China's Sinovac vaccine on P.1.
Whatever properties it has that have brought about its dominance — and potentially Manaus's second crisis — epidemiologists are alarmed.
"If it happened in Manaus, we need to be ready for the possibility of it happening elsewhere," says Hanage. "There are large parts of the globe where there's not going to be significant vaccine induced immunity for some time."
P.1 has now been recorded across Brazil and in at least five other countries, including the USA. Some nations like the UK, which is struggling with its own highly transmissible variant, have limited travel from Brazil to reduce transmission of P.1.
In Manaus, it is too late to prevent its spread. Orellana is already working hard to prevent a third crisis, petitioning the Inter-American Commission on Human Rights for support instead.
But on February 6, state governor Wilson Lima loosened restrictions once again, even while deaths exceeded 140 a day.
"It's like being punched twice by the same person, forgiving them, and saying 'hit me a third time!'" Orellana says.
"It's madness. We don't want to be hit a third time."