In this paper we show unambiguously that ordering in ${\mathrm{Si}}_{1\mathrm{\ensuremath{-}}\mathit{x}}$${\mathrm{Ge}}_{\mathit{x}}$ is an entirely kinetic phenomenon governed completely by growth conditions and surface reconstructions, and not by bulk thermodynamic equilibrium. We describe a set of experiments involving the use of annealing, surface modification through molecular-beam-epitaxial adlayers, ultrahigh-vacuum chemical vapor-deposition growth, and growth on Si substrates with different orientations which indicate that both low-temperature growth and a 2\ifmmode\times\else\texttimes\fi{}1 reconstructed surface are individually necessary but not sufficient conditions to observe long-range order in ${\mathrm{Si}}_{1\mathrm{\ensuremath{-}}\mathit{x}}$${\mathrm{Ge}}_{\mathit{x}}$. The experimentally observed ordered phase, which consists of bilayers of Si and Ge along all four 〈111〉 directions, is a metastable phase that does not correspond to the lowest-energy phase of the alloy and is irreversibly destroyed by annealing. Our experiments also emphasize that ordering in ${\mathrm{Si}}_{1\mathrm{\ensuremath{-}}\mathit{x}}$${\mathrm{Ge}}_{\mathit{x}}$ occurs even in the absence of strain in ${\mathrm{Si}}_{1\mathrm{\ensuremath{-}}\mathit{x}}$${\mathrm{Ge}}_{\mathit{x}}$ films. Finally, we demonstrate that long-range order in ${\mathrm{Si}}_{1\mathrm{\ensuremath{-}}\mathit{x}}$${\mathrm{Ge}}_{\mathit{x}}$ occurs due to local segregation induced by stresses at the growing surface. This mechanism successfully explains all experimental findings to date.