Exact exponential rates for the discrete Calderon problem on lattice strips, and what transient data do not change

Exponential ill-conditioning of the inverse conductivity problem and of its discrete network versions is classical. This preprint adds a sharp exponent on a solvable family: a lattice strip periodic along the measured boundary. Translation invariance makes the Jacobian of the Dirichlet-to-Neumann map exactly block-diagonal in the total momentum along the boundary; each block is Vandermonde-type, and the smallest singular value of block q decays with depth at the exponent of the Green function of the complement of the node set on the unit circle (a Bernstein-Walsh argument, a sketch and not a proof). The rate is checked in 140- to 220-digit arithmetic on four geometries, including values fixed before the computation (aligned square strip 1.653 decades per row at the zigzag momentum; boundary along the lattice diagonal 1.067; lateral-to-vertical conductance ratios 0.05 and 16: measured 1.762 and 2.498 against predicted 1.765 and 2.451). A preregistered experiment with the CVODE integrator of rusty-SUNDIALS asks whether transient data, represented by real Laplace variables, change the depth growth of the condition number: they lower it by about one decade and leave the exponential rate unchanged (post hoc slope ratio 0.98); one prediction was refuted narrowly, and the time-to-frequency gate failed as written because of a design error of mine and passes only in a post hoc form. Four elementary linear-algebra facts behind the invisibility of constant offsets are machine-checked in Lean 4. A literature review (inverse conductivity, Vandermonde and Hankel conditioning, hyperbolic circuits) positions the work; closely related rigorous results exist for real nodes and Hankel matrices, and the novelty of the exact statement is not established. A register of every failed gate and refuted prediction is included. Nothing here concerns holography. Companion to doi:10.5281/zenodo.23228685.