After reviewing the experimental
data for $v_n$, we argue that several features observed across
energies and system sizes show fundamental tensions with the
hydrodynamic model, making the search for a non-hydrodynamic
explanation a worthwhile alternative. We then show that the GLR
equation develops unstable modes when the parton distribution
function is generalized to depend on
azimuthal angle. This generally means the saturation scale $Q_s$
aquires an azimuthal dependence. We argue that this process is a
plausible alternative explanation for the origin of elliptic flow,
one that naturally respects the scaling experimentally observed.
Nuclear
Physics Colloquium