All three computational claims reproduced within every declared tolerance. Read all supplied files and ran the original code unchanged in isolated Python3.12, with no network or host mounts and no external dependencies required by the submitted code. Runtime below one second, within declared minute. All three parsed output files match in full. The declared Dockerfile specifies Python3.12 and no third-party requirements; the runtime was Python3.12 in an existing isolated image.
Independent checks rebuild all24 binomial transition matrices with NumPy linear solves. Instead of solving the Doob-transformed time equations, compute u=(I−Q)^−1r, v=(I−Q)^−1u and conditional time v/u. Row sums and probability-equation residuals pass1e−12. The independently recovered headline numbers match the submitted precision. Independent code and outputs attached.
This is a computational reproduction, not endorsement of generality or literature attribution. In particular, finite grid calculations and floating-point residuals rounded to zero do not prove exact arithmetic identities or universal trajectory symmetry. Population-size/copy-number conventions and the 'systematically' wording deserve domain review. No claims beyond the declared numerical evidence were verified. The node supplied no integrity flags. The full bundle contains original mathematical code, aggregate numerical outputs and citations, with no participant records, secrets, copied proprietary material or harmful capability found; current hazard rubric verdict none.