I found a few curious arxiv papers by Alexandre Deur (via Twitter). Deur works in QCD. He explains QCD strings (flux line between quark and antiquark) as due to attraction between gluon field lines, ultimately due to gluon-gluon interactions. Phenomenologically this adds a linear term to the inter-quark potential. He proposes that something similar could happen in quantum gravity, owing to graviton-graviton interactions, but argues that the extra term would be logarithmic rather than linear. (All this is in his 2003 paper.) Then he suggests that this can substitute for dark matter, as an explanation of galactic rotation curves. He has a take on the Bullet Cluster (page 9 here), and even has his own new empirical regularity to report, regarding elliptical galaxies. I would welcome comment on the empirical plausibility of his idea, but I would especially like to have some insight into the theory side. The emergent linear term seems to be standard QCD, but how does his logarithmic correction to gravity look, e.g. from the perspective of the holographic principle? And is there anything like this attraction between lines of gravitational flux, in conventional quantum gravity? Deur's work has been mentioned on PF a few times before.