Neither. Because neither is wrong. And the article is trying to sensationalize a claim the scientists didn't make.
It is the average speed of the light over very large distances that needs a correction, to account for the portions of travel where the light, well, is not light. The photons still move at 2.99x10^8m/s. It's the electrons and positrons that move slower.
This is incorrect, virtual particles don't move slower, or move at all. They are probability waves that appear in particular space-time point and collapse at another. That's why they can annihilate again, at all; if they moved as normal particles, they should have flown at opposite directions due to conservation of momentum and never see each other again (think about that). This all happens actually at the speed of light. I feel like the paper's authors mess up with things they don't really understand.
None of this is the issue; speed of light stays constant, as does distance measurements. What changes is the understanding of the stability of a photon of light in a vacuum and the effect of this instability on travel time while passing near a gravitational well.
So while it's a photon of light, it travels light speed. When the energy converts to kinetic energy for a breather, it is affected by the gravitational pull, in a manner significantly stronger than a neutrino is affected. When it then flops back to being a photon, it is once again traveling at the speed of light.
What intrigues me about this is that this will also have implications regarding relativity, as every time the light flips state, it is essentially anchoring itself to a location in space from which the next photon flop can take its bearing. My mind can't quite grasp the further implications of this right now, but it could really mess with observation of light from a moving point (which all points are).
The recalibration is mostly on how we project distances based on light measurements; it's now become significantly trickier, as we need to account for gravity at specific moments.
Gravity affects energy, not mass. Regardless the photon travels as a light or as a pair of virtual electron and positron, the gravity effect is the same. Why otherwise do we have gravitational lenses. I don't really know what this article is all about; gravity and quantum theory don't quite mesh well, it is well known. This sort of result, even assuming it is conceptually and mathematically solid, which is a far stretch, is a poke in the sky.
When light travels through a medium containing matter it will be absorbed and "stored", for some time, in the exited states of the atoms before it is emitted again. This can account for the "slowing" down effect, which again means that the photon you measure with your photon detector wasn't actually slowed down, it simply not the same photon that was emitted from your photon source.
THIS
This is the source of all troubles in the world - bringing in improper analogies and building "theories" upon them. Let's begin with that the photon is not a particle flying through space, it is a probability wave. Particles do not fly, the probability wave propagates. Likewise, light remains a wave in medium, and it is that wave that is slowed due to interference with the charges in atoms.
Try to explain with your little analogy the greater speed of light in some meta-materials; what, the atoms emit photon before they absorb it?
Some times it is just better to accept that your thinking platform needs a change than try to continue squeezing the world into a completely inadequate concept-box.
A penny saved is a penny to squander. -- Ambrose Bierce