OCE reformulates editing-based concept erasure in diffusion models as multiplicative parameter updates derived from closed-form orthogonal transformations. Empirical analysis shows concept semantics depend primarily on neuron direction while generative capacity relies on angular geometry (arXiv:2605.28902). Additive updates in prior methods entangle these factors and introduce interference. OCE applies layer-wise transformations that preserve neuron magnitude and geometry, achieving erasure of up to 100 concepts in 4.3 seconds with improved non-target preservation. For multi-concept cases it introduces subspace-level objectives and structured manipulation to resolve conflicting constraints. Related editing approaches in Gandikota et al. (arXiv:2303.07345) and Kumari et al. (arXiv:2304.05719) demonstrated similar scalability limits under additive regimes.