The geological evidence suggests that there have been liquid oceans on Earth dating back at least 3.5 billion years, a fact that was crucial to the emergence of the very first lifeforms. The only problem is that that's utterly impossible.
Here's the basic problem. In the first billion years or so of the Earth's existence, the Sun would have been considerably dimmer than it is today, perhaps only about 70% of its current brightness. In those days, the Sun wouldn't have given off enough heat to keep liquid water on Earth's surface without it freezing back up. But that's contradicted by ancient rocks that show clear signs of liquid water dating back that far, not to mention the fact that the earliest ancestors of life apparently began to emerge way back then, and they almost certainly would have needed liquid oceans to do that.
There are a few possible explanations for this little mystery. The Earth might have had a lower albedo back then, meaning that it reflected less of its heat back towards the Sun. The most obvious way to lower the planet's albedo would be to substantially reduce its cloud cover. Another possibility is that Earth might have suffered from a runaway greenhouse gas effect, a little bit like what we find on Venus today. Either way, the mystery was big enough to warrant this rather impressive title: "The Faint Young Sun Paradox."
About a year ago, researchers announced a possible solution to the paradox. They argued that the rocks dating back to that time period showed no evidence of methane or carbon dioxide, two of the most common greenhouse gases. That made the lower albedo theory the clear favorite, and the researchers argued that in the absence of lots of biological particles that would nucleate water droplets, fewer clouds would have formed.
Here's the problem though - all the researchers did was provide evidence against the greenhouse effect theory, not offer evidence for the lower albedo theory. Researchers Colin Goldlatt and Kevin Zahnle at NASA's Ames Research Center have now approached the question from the other direction, assessing the logistics of the lower albedo theory. And it really doesn't look good.
They discovered that no matter how you arrange the clouds, there's no way that it could have lowered the albedo enough to keep the oceans liquid. According to repeated simulations, the most that Goldlatt and Zahnle could get was about half the necessary heating to keep the water unfrozen. Specifically, since higher clouds trap heat while lower clouds reflect it, they tried a simulation where there were no lower clouds at all, and even so they say that "even with the strongest plausible assumptions, reducing cloud and surface albedos falls short by a factor of two of resolving the paradox."
The two best theories to explain why the ancient Earth didn't freeze have now seemingly been disproved. There's no easy way out of this paradox, and we may need an entirely new theory to account for it, because this is one of those times where the theoretical and practical evidence seem hellbent on disagreement.