We investigate the synthesis of (In,Ga)N/GaN multiple quantum wells by plasma-assisted molecular beam epitaxy. For metal-stable growth, reflection high-energy electron diffraction and X-ray diffraction reveal massive In surface segregation which is directly confirmed by In depth profiles recorded by secondary-ion mass-spectrometry. These profiles exhibit a top-hat In distribution and are thus indicative of a zero order segregation mechanism instead of a first order process as observed for other materials systems. The segregation of In during metal-stable growth results in quantum wells with smooth interfaces but larger width than intended, and thus causes blue-shifted transition energies and poor quantum efficiency. This unexpected blue-shift may be the reason for the frequent conclusion that the theoretical polarization fields of Bernardini et al. [Phys. Rev. B 56, R10024 (1997)] are too large for (In,Ga)N. Being in possession of the (at least approximately) correct structural parameters, we find the theoretical fields of Bernardini et al. for (In,Ga)N to be in very satisfactory agreement with the experimental data. Reduction of In segregation by N-stable conditions is possible but inevitably results in rough interfaces.