TDA manuscript: deterministic stress-study reproduction map

seed isn't the whole reproduction recipe. draw order + versions + tolerances + finite-bar conventions too, otherwise same seed / different diagrams somehow

Protocol distinction: the supplement's 300 cloud-pair stress study uses a common ruler of one, not a separate empirical quantile for each cloud. Do not mix this experiment with the separate-ruler obstruction example.

Recorded environment: Python 3.10.19, NumPy 2.2.6, SciPy 1.15.3 and GUDHI 3.11.0. Rips complexes are built through dimension k+1 over F11; finite degree-k intervals are retained and the GUDHI bottleneck routine is called with e=0.

Reference code path: analysis/independent_evidence_checks.py. The supplement describes an independent sparse-boundary reduction and bottleneck assignment, checked against GUDHI. Agreement to displayed precision is explicitly distinguished from binary-identical output.

Reported table: 150 H0 pairs and 150 H1 pairs; median conservative-baseline/new-bound factors 3.59 and 2.42, respectively; observed total factor range 1.26–10.69 and no computed exceedance of the bound. These are results reported for one deterministic protocol, not fresh measurements or estimates of typical application performance.

Review task: recover the exact random-draw order, record file and tolerances from the archive before rerunning. A seed alone is insufficient if array operations or draw ordering change.