Superconductivity in iron selenides has experienced a rapid growth, but not without major inconsistencies in the reported properties. For alkali-intercalated iron selenides, even the structure of the superconducting phase is a subject of debate, in part because the onset of superconductivity is affected much more delicately by stoichiometry and preparation than in cuprate or pnictide superconductors. If high-quality, pure, superconducting intercalated iron selenides are ever to be made, the intertwined physics and chemistry must be explained by systematic studies of how these materials form and by and identifying the many coexisting phases. To that end, we prepared pure K2Fe4Se5 powder and superconductors in the KxFe2-ySe2 system, and examined differences in their structures by high-resolution synchrotron and single-crystal x-ray diffraction. We found four distinct phases: semiconducting K2Fe4Se5, a metallic superconducting phase KxFe2Se2 with x ranging from 0.38 to 0.58, the phase KFe1.6Se2 with full K occupancy and no Fe vacancy ordering, and a oxidized phase K0.51(5)Fe0.70(2) Se that forms the PbClF structure upon exposure to moisture. We find that the vacancy-ordered phase K2Fe4Se5 does not become superconducting by doping, but the distinct iron-rich minority phase KxFe2Se2 precipitates from single crystals upon cooling from above the vacancy ordering temperature. This coexistence of separate metallic and semiconducting phases explains a broad maximum in resistivity around 100 K. Further studies to understand the solubility of excess Fe in the KxFe2-ySe2 structure will shed light on the maximum fraction of superconducting KxFe2Se2 that can be obtained by solid state synthesis.