Rerun of release/reproducibility from an extracted copy of release/source.zip (version of 2026-10-02 20:39,
104,015 bytes), verification run 2, 2026-10-02.  Extracted into a scratch directory; nothing was written into the
package.  Compiler: Apple clang 21 (c++ -O2 -std=c++17, -pthread where the README says so); Python 3.13.

Build: enum_conj2.cpp, enum_conj3.cpp, cyc_class.cpp (with -pthread), thm1_bruteforce.cpp, conj3_bruteforce.cpp
(pruned and -DNOPRUNE) all compiled without errors from the shipped instructions.

Python scripts rerun (output compared byte for byte with reproducibility/outputs/):
  verify_paper_examples.py              identical
  conj2_all7.py                         identical
  multipath_examples.py                 identical
  threepath_example.py                  identical
  c3_B17_proof_check.py                 identical
  c3_example.py B17 / A17               identical / identical
  count_linear_extensions.py            identical
  conj2_exhaustive.py 11                identical
  thm2_check.py 60000 1                 identical
  independent/t_thm2.py 1 80000         identical
  test_observation_A.py 40000 1         identical
  test_observation_B.py 20000 1         identical
  apparent_pairs_check.py 20000 1       identical
  emit_classes.py 17                    "28 classes written to classes" (identical)
C++ programs rerun:
  enum_conj2 11                         identical rows n <= 11
  enum_conj2 13                         identical to the saved output, i.e. 709593014 filtrations at n = 13 (see below)
  enum_conj3 11                         identical
  conj3_pruned B17 / A17                identical / identical
  cyc_class on the 17 classes of size <= 16 (6 threads)   identical to cyc_class_le16_output.txt
  thm1_bruteforce on classes 00-08      identical lines
  thm1_bruteforce on inputs/control17.txt  identical (1412014592 filtrations, 437696 with a closed V-path)
Not rerun: thm1_minsearch.py 17 (30 min), the 12-part Python run of Prop. 5.3, the size-17 cyc_class run and the
unpruned B17 run (each of these was reproduced by the independent programs of run 2 instead).

FINDING: enum_conj2.cpp caps the number of vertices at 12 (line 97: "if(nv<12) cand[nc++]= (u16)(1<<nv);", arrays
posOf[1<<12]).  It therefore misses exactly one filtration with 13 simplices, the one consisting of 13 vertices and
nothing else (every other filtration with 13 simplices has an edge and at most 12 vertices).  The correct count at
n = 13 is 709,593,015 (r2_enum.cpp); all other columns are unaffected because that filtration has no pair of
positive dimension.
