The three findings
1. Their SWMM5 results are real. EXTRAN is exact at Δt = 5 s, diverges by 120 s, and the Preissmann SLOT is genuinely unstable even at 5 s. All three reproduce here. This is not a dispute about their measurements.
2. It's explicit-vs-implicit, not engine quality. SWMM5's solver is explicit → it carries a Courant limit (≈ 6–16 s for 100-ft pipes). FSGI is implicit → unconditionally stable by construction. Running the explicit engine at a fixed 120 s step and calling the result an engine verdict is tautological.
3. Mass balance proves nothing here. Every curve — including a hydrograph diverging past 6,000 cfs — reports 0.00% continuity, because the oscillation is symmetric. The second white paper's own conclusion concedes this, yet the first paper's evidence is built on continuity tables.
Where to look
TEST 1 sweep — their exact flat-pipe case with a Δt slider. Watch the blue curve leave the truth while continuity holds at zero.
Stability cliff — 36 engine runs mapping precisely where the explicit scheme breaks, and why SLOT never gets a stable window.
Numerical diffusion — why "SWMM6 distorts the hydrograph" describes something every solver does at a large step.
Continuity audit — peak error plotted against mass-balance error. The correlation is nil. That chart is the whole argument in one picture.
Claims ledger — each published claim marked verified, reframed, or unsupported. Reproduce — run every number yourself.
Scope, stated honestly: only the SWMM5 side is reproduced here, because the EPA engine is public. FSGI and the real SWMM6 alpha binaries were not available, so nothing here measures those — claims about them are read from the papers, not tested.