In the parton model for deep inelastic scattering, the momenta of the partons are assumed to be longitudinal and on-shell. What's peculiar (at least rhetorically) is that even the "virtual" sea quarks are assumed to be on-shell. How does QCD justify these assumptions? Are there systematic corrections coming from non-vanishing transverse momentum and off-shell partons? Or are such corrections vanishingly small in some appropriate limit?