Alloy systems such as ${\text{Ga}}_{1\ensuremath{-}x}{\text{In}}_{x}\text{As}$ consist of different random assignments $\ensuremath{\sigma}$ of the Ga and In atoms onto the cation sublattice; each configuration $\ensuremath{\sigma}$ having, in principle, distinct physical properties. In infinitely large bulk samples different $\ensuremath{\sigma}$'s get self-averaged. However, in finite quantum dots (QDs) $(\ensuremath{\le}{10}^{5}\text{ }\text{atoms})$, self-averaging of such configuration $\ensuremath{\sigma}$ may not be complete, so single-dot spectroscopy might observe atomic-scale alloy randomness effects. We examine theoretically the effect of such atomic-scale alloy randomness on the fine structure-splitting (FSS) of the multiexciton observed via the polarization anisotropy of its components. We find that (i) The FSS of the neutral monoexciton ${X}^{0}$ changes by more than a factor of 7 with $\ensuremath{\sigma}$. Thus, dots provide clear evidence for the effect of the atomic-scale alloy randomness on the optical properties. (ii) For multiexcitons, the effect of alloy randomness can be so large that the polarization of given emission lines in samples that differ only in random realizations can be dramatically different, so it cannot be said that given transitions have fixed polarization. (iii) Polarization is affected both by atomic-scale randomness and by possible geometric elongation of the QD in one direction. Because of different random realizations, even 50% QD base elongation in [100] direction gives the same polarization as in a geometrically symmetric dot. Thus, measured polarization cannot be used to determine QD elongation.