The moon is the most obvious and familiar object in Earth’s night sky — constant, consistent, predictable in its monthly cycles and its daily rising and setting. Astronomers understand the moon’s movements so thoroughly that even a break from the routine, like an eclipse, can be anticipated 1,000 years in advance.

But we don’t know the moon as well as we think. In fact, for years, astronomy has been in an uproar over the origin of Earth’s only natural satellite, grappling to make sense of a model that seems increasingly unsatisfactory.

Now, a team of Israeli researchers has shaken up the debate by offering an entirely new explanation, published this week in the journal Nature Geoscience. They say the moon isn’t a single chunk of rock but an amalgamation of nearly two dozen “moonlets,” one that was formed during a steady bombardment of Earth by several smaller bodies.

It’s a major departure from the “giant impact model,” which was once the standard explanation for the moon’s existence. That hypothesis proposes that the satellite came about during a single, violent collision between Earth and a hypothetical protoplanet called Theia. Theia sideswiped our planet roughly 4.4 billion years ago, scattering debris that eventually coalesced into the moon, which drifted away and started to circle the Earth.

The model explained the moon’s modern migration — it’s still receding, at a rate of about 4 centimeters per year. But it also had a problem: In the early 2000s, scientists examining lunar rocks brought back by the Apollo astronauts found the chemical composition of the moon was eerily similar to that of Earth. Its elements had the same ratio of isotopes as many on Earth do — and virtually no traces of Theia. How could a giant object create the moon and leave nothing behind?

“The whole giant impact model had been put into crisis several years ago,” Sarah Stewart, a planetary physicist at the University of California at Davis, told The Washington Post last year, “to the point where people thought it might be completely wrong because we couldn’t make it work in its details.”

But every tweak seemed to make the giant impact model even more improbable.

“If you do that too many times, everyone in the room starts squirming,” Stewart said.

A more likely scenario, researchers argue in the new Nature Geoscience paper, is a series of smaller impacts. Lead author Raluca Rufu, a planetary scientist at the Weizmann Institute of Science in Israel, knew that the solar system’s infancy was a chaotic time. Small bodies (a 10th the size of Earth or less) ran rampant in the system, bumping into things like rambunctious toddlers. Though a collision with a major protoplanet like Theia would have been rare, bombardment by these smaller bodies happened frequently.

Each collision would have sent a spray of debris into orbit around Earth, forming disks made mostly of material from Earth (rather than the impactor). Each disk then cooled and coalesced into a moonlet, which would migrate outward and glom onto other newborn small rocks, forming a growing moon. About 20 of these moonlets could have combined to create the satellite that orbits Earth today.

This version of the moon’s origin story solves the big chemistry question that dogs proponents of the giant impact model. Creating a moon from many impacts dilutes the influence of each impactor on its chemistry — their individual signatures would have been lost amid all the Earth material each moonlet carried. That could explain why lunar samples are more like Earth than anything else in the solar system.

The idea for a many-impact model has been around since the 1980s, but no one was sure if such a process could create sufficiently large moonlets. With more than 1,000 computer simulations of potential ancient impacts, Rufu and her colleagues demonstrated that sizable moonlets are a fairly common outcome of these crashes.

Proving the second part of the hypothesis — that moonlets would bunch together, rather than drifting away or getting reabsorbed by the Earth — is Rufu’s next challenge. Right now, she and her colleagues have no model for that merging process.