Referee 2 cross-check log (2026-09-30)

1. Independent code written before reading any finder/lead/verifier script:
   - ref2_checks.py 3000 22      -> ref2_checks_3000_22_output.txt       (all PASS, 63 s)
   - ref2_derivpoly.c (derivative polynomials of tan and sec; no triangle, no convolution)
       ref2_derivpoly 131232 16  -> ref2_derivpoly_131232_16_output.txt  (PASS, k <= 16)
       ref2_derivpoly 1048896 19 -> ref2_derivpoly_1048896_19_output.txt (PASS, k <= 19, 12m45s)
   - ref2_oddprime_side.py       -> ref2_oddprime_side_output.txt        (side check of the Gulec value s(5^5)=4, d(5^5)=2500 quoted in Remark 6.2)
   Build: cc -O2 -o ref2_derivpoly ref2_derivpoly.c

2. Bitwise cross-check of two independent methods:
   finder euler_mod2_64.c (Seidel-Entringer-Arnold triangle) run for N = 131232, and
   ref2_derivpoly 131232 (derivative polynomials), both writing E_0..E_N mod 2^64 as uint64:
   the two binary files are byte-identical (cmp: no difference) for all n <= 131232.

3. Reruns of the release scripts (copies in the scratchpad; release folder untouched):
   - lead/verify_ramassamy.py 1000 14   -> rerun_lead_verify_ramassamy_1000_14_output.txt : 18 checks, 0 failures, 24.6 s;
     identical to release/reproducibility/lead/verify_ramassamy_output.txt except the timing lines.
   - referee/verifier1/v1_indep_triangle.py 3000 -> identical to the stored output (except timing lines)
   - referee/verifier1/v1_indep_convolution.py 16 -> identical to the stored output (except timing lines)
   - referee/verifier2/my_triangle.py 3000 -> identical
   - referee/verifier2/my_ftransform.py 22 -> identical

4. Paper build: a scratch copy of paper/main.tex + references.bib builds with tectonic --keep-logs with no errors,
   no warnings and no overfull boxes (11 pages); all 11 pages of release/paper.pdf were rendered at 1.4x and inspected.
