Inhibitor segregation can make a nonexplosive reaction network explode

Unrefereed research candidate. A stochastic chemical system explodes when infinitely many reaction events occur in finite time. Anderson, Howells and Rojas La Luz (2026) conjectured that a nonexplosive chemistry stays nonexplosive inside dividing compartments and pointed to inflow reactions as the main danger. This package gives two two-species chemistries with no inflow reaction, no compartment fusion and no exit that are nonexplosive alone but explode with positive probability when compartments divide at rate λs, keep all substrate in one daughter and retain each inhibitor with probability Q (fresh at each division). For conversion inhibition (2S→3S, S→S+I, S+I→2I) the sharp condition is d+λE[log Q]<b; for catalytic inhibition (I+2S→I+S) it is bλ(1−EQ)>dc; both boundaries are nonexplosive. Also proved: exact conditional explosion laws, blow-up asymptotics, monotonicity and continuity at threshold, robustness to bounded substrate leakage, and a comparison with well-mixed degradation showing a strictly larger explosive region for bulk export of a self-reproducing inhibitor and the same threshold for a catalytic one. For a three-molecule example the explosion probability is at least 1/2 and lies in [0.79744, 0.79751]. The archive contains the 24-page manuscript (PDF and LaTeX), exact rational checks, a Monte Carlo estimator, a two-sided enclosure program, a replay runner, semantic negative controls, CI and SHA-256 sums. Assurance: written proofs and computational checks; internal AI-assisted review only (not peer review, independent reproduction or formal verification). The probabilistic components are classical and cited; Yoo (2026) reported an earlier counterexample to the unrestricted conjecture using inflow and fusion. Creator: Anonymous. Text and data CC0-1.0; code MIT.