Roughly 650 million years ago vast sheets of glaciers stretched from the poles to the tropics, entombing Earth within a frozen skin that lingered for millions of years. And this had happened before: Our “pale blue dot” has transformed into a pearly-white “snowball Earth” at least three times in our planet’s history. But these deep freezes present a conundrum: They should have been deadly and yet life clearly survived. There is both geologic evidence our earliest microscopic ancestors did not freeze to death and genetic indications the lineages of a range of single-celled organisms extend beyond snowball Earth. The question is how.

A new study published to the preprint server arXiv and submitted to Earth and Planetary Science Letters might provide a resolution. Adiv Paradise, an astronomy graduate student at the University of Toronto (U.T.), and his colleagues modeled a variety of possible snowball worlds—varying the numbers of volcanoes they host and the amount of stellar light they receive—only to find many of these worlds would never escape snowball status. Those that had little volcanic activity would never emit enough carbon dioxide to spark the runaway global warming needed to wake them from their cryogenic slumber (as likely happened on Earth). Yet surprisingly, many of these worlds could also support large unfrozen pockets of land. Some of those areas remain dry, like the McMurdo Dry Valleys in Antarctica, but others develop local hydrological cycles, allowing liquid water to pool and flow across their surfaces.

Such oases are one explanation for how snowball worlds might remain habitable—a result that could describe not just Earth but many of the planets astronomers are discovering across the galaxy. “Before we might have brushed a snowball off as not being habitable, and we would have missed that there could be pockets of life,” says co-author Diana Valencia, an astrophysicist at U.T.

Indeed, the study aligns with previous work on the most recent freezing episode in Earth’s history. In 2015 Douglas Benn, a glaciologist at the University of Saint Andrews in Scotland, published a study that shows Earth’s climate was sensitive to variations in our planet’s orbit around the sun, resulting in cycles of ice sheet advance and retreat. The latter allowed lakes to pool, rivers to flow and simple microbial life to flourish—even during a snowball event. Benn and his colleagues saw such cycles in computer models they created of Earth’s climate and they also found sedimentary deposits in the Arctic Ocean islands of Svalbard that preserve evidence for the advance and retreat of the ice sheets. The findings imply the last snowball Earth would not have been a total “deep freeze”—that ice-free pockets of land existed where water could flow—thus sustaining a crucial refuge where life could persist until more favorable conditions returned.

But ice-free zones are not the only mechanism proposed to explain how life survived on snowball Earth. Since 1992 researchers have hypothesized an array of ideas, and every scientist appears to favor a different one, says James Kasting, a geologist at The Pennsylvania State University. He has argued life might endure below a thin layer of ice. In Antarctica lakes freeze so slowly that they do not include air bubbles and thus remain transparent to sunlight—allowing photosynthetic life to thrive beneath several meters of ice. And Paul Hoffman, a retired geologist from Harvard University, argues dust might provide the most likely reprieve for life. As snow collects dust it can more readily absorb sunlight, causing ponds of meltwater to form on the ice. Such ponds are well known in polar environments today to host thriving ecosystems of algae and cyanobacteria (although Benn notes scientists have no direct geologic evidence of these ponds at the time of snowball Earth). Finally, no geologist argues against hydrothermal vents, where volcanically active areas spew water at superhot temperatures. Hot springs in Antarctica and Iceland, after all, create warm oases that ooze with life today.

Ultimately, the jury is still out on which mechanism helped life pull through snowball Earth. Although Kasting notes that the ice-free zones hypothesized by both Paradise and Benn provide one potential solution, there are several caveats to the latest model. Both he and Hoffman would like to see Paradise’s team include sea glacier flow, for example, because it is possible that ice could flow from the poles to the equator, covering the nonglaciated areas they propose. And Paradise himself lists an array of caveats for his model: it is low in resolution, took a few computing shortcuts and does not include certain processes like the effects of atmospheric dust.

At the end of the day, there might be yet another survival mechanism that no one has thought of yet, Kasting says. Or it could also be several mechanisms worked together to help life endure here on Earth. Benn argues life likely did not survive in one major environment, but multiple environments. As such, snowballs might remain habitable with the help of ice-free zones, thin ice sheets, ponds of meltwater and hydrothermal vents. Indeed, Joseph Kirschvink, a geobiologist at the California Institute of Technology who coined the phrase “snowball Earth,” has always been surprised that so many people expected life to vanish within the deep freeze. “Life is hard to extinguish—even on a snowball,” he says.