RERUN_LOG.txt -- re-runs of 11 October 2026 (Python 3.13.5, sympy 1.14.0, mpmath 1.3.0, numpy 2.5.2, macOS, one
laptop)

Part 1.  Output of `sh run_quick.sh`, run by the third verification run from an extracted copy of the final
archive (before this log was added; total time 9 minutes 49 seconds, one process, on a laptop that was running
other jobs at the same time).  The same run had been made when the note was written, with the same result.
"passed": the program ended with "ALL CHECKS PASSED" (exit status 0).  "identical": the new output is
byte-identical to the recorded file.  "identical up to times": identical after replacing the fields that record
running times ("12.3 s").  For check_certificates.py the comparison is made block by block, for the blocks of the
seven parameter points that the quick run repeats.

== writing_stage/ (programs written with the note)
passed                check_certificates.py, parts A-D, for N1000,N700-B,N700-A,N690-A,N680-A,control-N3-source-example,control-N4-three-steady-states
identical up to times check_certificates_output.txt, block of N1000
identical up to times check_certificates_output.txt, block of N700-B
identical up to times check_certificates_output.txt, block of N700-A
identical up to times check_certificates_output.txt, block of N690-A
identical up to times check_certificates_output.txt, block of N680-A
identical up to times check_certificates_output.txt, block of control-N3-source-example
identical up to times check_certificates_output.txt, block of control-N4-three-steady-states
identical             writing_stage/rejection_tests_output.txt
identical up to times writing_stage/check_spectrum_output.txt
identical             writing_stage/check_spectrum_selftest_output.txt
identical             writing_stage/tables_and_figure.txt
identical up to times cross_checks/b4_on_pointA_N700.out
== original/ (programs with which the results were first obtained)
identical up to times original certify_five_steady_states.out
identical             original small_N_checks.out
== verification_run_A/
identical up to times A a2_theoremA_exact.out
identical up to times A a5_exactly_five_vca_N1000.out
identical             A a5_selftest.out
identical up to times A a4_propB_small_N.out
== verification_run_B/
identical up to times B b2_exact_ivt_N1000.out
identical up to times B b2_exact_ivt_N2000.out
identical up to times B b4_exact_count_N1000.out
identical up to times B b4_exact_count_N2000.out
identical up to times B b4_exact_count_PAPER3.out
identical up to times B b4_exact_count_N660.out
identical up to times B b4_exact_count_N700.out
identical             B b10_crosscheck.out
identical up to times B b6b_spectrum_state1.out
identical up to times B b6b_spectrum_state2.out
== number of differences: 0

The run wrote 26 files into the new folder rerun_out/ and changed no other file of the extracted copy
(checked by comparing every file with the archive).

Part 2.  The complete run of writing_stage/check_certificates.py (all fifteen parameter points and the two
controls, parts A to D), recorded in writing_stage/check_certificates_output.txt:
"17 parameter points checked in 3441 s:  ALL CHECKS PASSED" (exit status 0).  Running times per point, as
recorded there: N = 1000: 29 s; N = 2000: 227 s; the five points of run B (N = 651, 652, 655, 660, 700): 194, 92,
48, 32, 19 s; the eight points of run A (N = 652, 653, 655, 660, 680, 690, 700, 4000): 82, 65, 55, 38, 29, 28,
24, 2481 s; the two controls: below 0.1 s.  For N = 4000 the parts A to C alone take 31 s (separate run with
--only N4000-A --parts ABC); the rest is part D.
The third verification run repeated this complete run from an extracted copy of the archive as it was before
its revision (the checker, its data file and the recorded output are unchanged in the final archive), in four
processes (--only N4000-A; --only N2000,N651-B; the seven other points with N <= 660; the remaining points and
the two controls within run_quick.sh): ALL CHECKS PASSED in every process, and every block of the output is
identical to the recorded one up to running times (N = 4000: 2257 s).

Part 3.  Other recorded runs of 11 October 2026 in writing_stage/: rejection_tests.py (22 corruptions, all
rejected; exit status 0), check_spectrum.py ("total 49 s:  ALL FIVE STATES CERTIFIED"), check_spectrum.py
--selftest, window_widths.py, make_tables_and_figure.py, and the cross-checks of cross_checks/ (the programs
b4_exact_count.py and b2_exact_ivt.py of run B on the eight points of run A).  The re-runs of the programs of
original/, verification_run_A/ and verification_run_B/ that were made at the writing are listed in README.md,
section "Re-runs of 11 October 2026".  The third verification run also repeated b6b_rigorous_spectrum.py for the
steady states 3, 4 and 5 (identical up to running times).

Part 4.  The programs of the third verification run and their recorded outputs are in independent_run_2/
(see the README there).
