The achievement of universal quantum computing critically relies on scalability. However, ensuring the necessary uniformity for scalable silicon electron spin qubits poses a significant challenge due to the considerable fluctuations in valley splitting energy ($E_{\textrm{VS}}$) across quantum dot arrays, which impede the initialization of qubit systems comprising multiple spins and give rise to spin-valley entanglement resulting in the loss of spin information. These $E_{\textrm{VS}}$ fluctuations have been attributed to variations in the in-plane averaged alloy concentration along the confinement direction of Si/SiGe quantum wells. In this study, employing atomistic pseudopotential calculations, we unveil a significant spectrum of $E_{\textrm{VS}}$ even in the absence of such concentration fluctuations. This spectrum represents the lower limit of the wide range of $E_{\textrm{VS}}$ observed in numerous Si/SiGe quantum devices. By constructing simplified interface atomic step models, we analytically demonstrate that the lower bound of the $E_{\textrm{VS}}$ spread originates from the in-plane random distribution of Si and Ge atoms within SiGe barriers—an inherent characteristic that has been previously overlooked. Additionally, we propose an interface engineering approach to mitigate the in-plane randomness-induced fluctuations in $E_{\textrm{VS}}$ by inserting a few monolayers of pure Ge barrier at the Si/SiGe interface. Our findings provide valuable insights into the critical role of in-plane randomness in determining $E_{\textrm{VS}}$ in Si/SiGe quantum devices and offer reliable methods to enhance the feasibility of scalable Si-based spin qubits.