# ADDENDUM to PREREG_einstein_pipes — E3's fit window (30 September 2026) Run 1 crashed at mass 5, drag 1: the window "t > 3m/γ and ⟨cos θ⟩ > e⁻³" was empty. By t = 3τ = 15, ⟨cos θ⟩ had already fallen to about e⁻⁵. The defect is mine, in the window's design. Masses 0.2 and 1 measured 0.5068 and 0.4984 against 0.5 before the crash (run 1 output kept). Same quantity, same criterion (within 5% of T/γ): the rate is measured from the spread of the unwrapped angle, Var θ(t) = 2D(t − τ + τe^(−t/τ)), which is −2 ln⟨cos θ⟩ for Gaussian θ. Its slope over t ∈ [3τ, 3τ + 3/D] is 2D, with no noise floor. E1, E2 and E4 are unchanged.