I think that the way to approach physics is not to aim to invent beautiful theories, but to look for the data that shows the way. One of the sign-post papers that happened to influence me in this way was: Scarpa, Marconi and Gilmozzi, 2006, although I've read similarly clear ones by, eg: M. Milgrom, S. McGaugh & X. Hernandez.
Scarpa et al. make the point in their introduction that Newton's laws have never been tested at the tiny accelerations that exist at the edge of galaxies and that "deviations from Newtonian dynamics are always observed when, and only when, the gravitational acceleration falls below ~10^-10 m/s^2 as computed considering only baryons". They go on to state that it is agreed that dark matter cannot affect things on the small scale of globular clusters (dark matter haloes are large and smooth), so they looked at three globular clusters and showed that, indeed, below the critical acceleration (from the mass in the cluster) they deviate from Newton, just like larger galaxies. This suggests the presence of new physics rather than dark matter.
They also make the point that the external gravitational field, from the larger galaxy, acting on the globular clusters is above the critical acceleration, but the non-Newtonian behaviour is still seen. This points away from MoND where dynamics depend on the total acceleration, and points towards MiHsC where it is the local mutual accelerations that matter.
Scarpa et al. isn't perfect, only three globular clusters were analysed, but I'd like to express my appreciation to all those, like them, who risk unpopularity to base their conclusions on direct observations of nature, eg: M. Milgrom, S. McGaugh, X. Hernandez, J. Anderson, M. Tajmar, CERN (especially the OPERA team).... Observing new regimes is hugely risky, but is the only way to get to new physics.