Something strange and wonderful might be happening out in the border regions of galaxies: pairs of black holes merging again and again to produce phenomenal amounts of energy, more than a single star could possibly generate on its own.
That's the thinking of scientists studying dense clusters of stars - pockets of space where hundreds of thousands or even millions of stars congregate tightly together.
With these stellar objects in such close proximity, once two orbiting black holes merge, the resulting supermassive black hole could find a new partner, repeating the process again – that's the hypothesis that an international team of astronomers are now working from.
"We think these clusters formed with hundreds to thousands of black holes that rapidly sank down in the centre," says lead researcher Carl Rodriguez, an astrophysicist at MIT.
"These kinds of clusters are essentially factories for black hole binaries, where you've got so many black holes hanging out in a small region of space that two black holes could merge and produce a more massive black hole.
"Then that new black hole can find another companion and merge again."
It's an idea that Rodriguez and his colleagues have shown could work based on advanced maths models, run on the Quest supercomputer at Northwestern University in New York.
The computer chewed through 24 different simulations, covering clusters from 200,000 to 2 million stars in size, and running over a course of 12 billion years. That's a lot of number crunching, but it showed that "second-generation mergers" could indeed happen.
The special ingredient in the calculations was to go beyond Newton's theory of gravity – which holds in the vast majority of cases – to consider two black holes whizzing by each other very closely, as might be the case in these densely packed stellar clusters.
By adding in Einstein's theory of general relativity, gravitational waves and all, the researchers showed that black holes could merge together, rather than getting kicked out of the cluster (as they would under Newton's classical laws).
To get more evidence of the phenomenon, the team will need help from the gravitational wave detector at LIGO.
LIGO has already shown that stellar binary black holes exist, two dying stars joining in a dance into a massive merger that produces a burst of gravitational waves.
If mergers can go on to produce more mergers, then LIGO should be able to detect this too.
"If we wait long enough, then eventually LIGO will see something that could only have come from these star clusters, because it would be bigger than anything you could get from a single star," says Rodriguez.
These packed, spherical clusters of stars the researchers have been investigating appear in most galaxies – our own Milky Way holds about 200.
The final piece of the puzzle is to figure out whether black hole mergers can stick around long enough to create other mergers: depending on the spin of the black holes as they combine, the resulting gravitational wave energy might well propel the giant, combined black hole out of the cluster like a rocket.
Or would it? So far LIGO has only detected low rates of spin from black holes, suggesting these stellar objects aren't spinning as fast as scientists previously thought.
The lower rates of spin that LIGO has noted are another sign that maybe these combined mergers could be possible.
So, all eyes on LIGO again, to see if its measurements can back up the modelling that Quest has done. The thinking is that anything with a mass of between 50-130 solar masses would have to be formed by a multiple merger, rather than a single star.
"My co-authors and I have a bet against a couple people studying binary star formation that within the first 100 LIGO detections, LIGO will detect something within this upper mass gap," says Rodriguez.
"I get a nice bottle of wine if that happens to be true."
The research has been published in Physical Review Letters.