Cold collisionless dark matter is a great homogenizer: N-body relaxation drives halos of similar mass toward the same NFW-like density profile, so galaxies at fixed mass should show broadly similar inner rotation curves with modest scatter (Navarro et al. 1997). The prediction is structural, one of the cleanest the particle picture makes.
The curves refuse to be similar. At fixed mass, observed rotation curves span everything from slowly rising core-like profiles to steeply rising cusp-like ones, a diversity far beyond the simulated scatter (Oman et al. 2015). And the diversity coexists with its opposite: the Radial Acceleration Relation binds rotation speeds to the baryonic distribution with remarkably small scatter across the whole SPARC sample (McGaugh et al. 2016). The combination is the trap. A universal dark profile cannot produce the diversity; baryonic feedback tuned to produce the diversity cannot explain why the baryons alone then predict each curve through the RAR. The model needs the dark halo to be simultaneously universal enough for the RAR and variable enough for the diversity, and no feedback prescription delivers both from first principles.
The standing makes rotation curves, the original dark matter evidence, into one of the particle picture's sharpest problems. WALLABY and SKA HI surveys will expand the curve census by orders of magnitude, with environment and kinematic state measured well enough to test what the diversity actually correlates with.