Re-run of the release package from an extracted copy of release/source.zip (verification run 2, 2026-10-20).
Commands (run in the extracted reproducibility/ folder, nice -n 10):
  python3 -I check_note.py                  -> check_note_output.rerun.txt   (19.4 s)
  sh run_sims.sh                            -> sim_output.rerun.txt          (6.2 s)
  cd original; python3 -I check_drift.py 20000 -> check_drift_output.rerun.txt (63.4 s)
Comparison with the recorded outputs (diff):
  check_note_output.rerun.txt  vs reproducibility/check_note_output.txt          : identical (TOTAL FAILS 0)
  sim_output.rerun.txt         vs reproducibility/sim_output_seeded.txt          : identical
  check_drift_output.rerun.txt vs reproducibility/original/check_drift_output.txt: identical up to blank lines (TOTAL FAILS 0)
Seeds 1 and 2 (empty start) and seed 3 (start (5, 20000, 40000, 0)) give the lines of original/sim_output.txt
(checked by comparing the avgN/maxN/final values and the 13 snapshot lines).
