Imagine you're running a bathtub with two drains and three faucets. You can reroute water between the faucets all day — swap hoses, add splitters, throttle one and open another — but the water level in the tub at steady state depends only on the total flow in minus the total flow out. The internal plumbing is irrelevant to the final level. That's the core mechanism this paper proves for salt transport in the Atlantic Ocean. The committed claim: in any closed thermohaline box model that conserves salt exactly, the Indo-Pacific gateway (Indonesian Throughflow, Agulhas retroflection, warm-route fraction) cannot change the equilibrium structure of the Atlantic Meridional Overturning Circulation. Every internal exchange cancels algebraically. The salt delivered to the Atlantic is fixed entirely by the net Atlantic freshwater export — a single external forcing parameter. The gateway enters as a purely additive constant and cannot renormalize the salt-advection feedback that drives AMOC hysteresis. The authors verify this with a parameter sweep showing departures of order 10⁻¹¹, i.e., machine-epsilon-level confirmation. The model is a five-box system — North Atlantic, upper-limb, Indian, Pacific, deep — with explicit return-flow splitting between warm (Indian) and cold routes, plus Indonesian Throughflow and Agulhas branches. This is a direct descendant of the Stommel (1961) two-box thermohaline model and Rahmstorf's (1996) extensions, but with the Indo-Pacific gateway boxes added specifically to test whether they modify the bifurcation diagram. The answer is a clean no. The proof strategy is elegant: sum the steady-state salt budgets of all gateway reservoirs and watch everything that isn't the Atlantic freshwater export cancel. It's an exact conservation argument, not a numerical approximation. The second half introduces distributed advective memory via a gamma kernel replacing the discrete transit-time lag in the return flow. This is where the dynamics get interesting. The equilibria are unchanged (conservation still holds), but the transient stability picture shifts dramatically. The authors derive a closed-form threshold for oscillatory (Hopf) instability that depends only on the shape parameter of the gamma kernel. Broad memory — meaning salt anomalies arrive smeared over a range of transit times — is strongly stabilizing. A discrete lag (all salt arriving at exactly one time) is the least stable case. Crucially, the oscillatory instability always precedes the saddle-node (fold) bifurcation, meaning the AMOC would start oscillating before it collapses, and the gap between oscillation onset and collapse widens more than tenfold as memory sharpens. The integrity regime here is mathematical proof plus confirmatory numerical sweep — about as clean as conceptual ocean modeling gets. The conservation result is exact and algebraic; it doesn't depend on parameter choices, numerical resolution, or simulation fidelity. The memory-kernel results are closed-form analytical. The authors are honest that this is a reduced (box) model, not a GCM, and that the result speaks to equilibrium structure, not to the full transient dynamics of the real ocean. The comparison baseline is the existing Stommel-Rahmstorf family of thermohaline models, not a competing GCM simulation. What this means for the AMOC debate: gateway processes — which have been widely invoked as potential stabilizers or destabilizers of the overturning — are shown to be thermodynamically inert at equilibrium in any salt-conserving framework. They can shape how the system oscillates on its way to a new state, but they cannot move the tipping point itself. This is a constraint that any higher-resolution model claiming gateway influence on AMOC stability must now explain its way around. The practical milestone is whether this conservation constraint survives in models with diffusion, mixing, and non-steady freshwater forcing — the next rung on the complexity ladder. The obvious experiment not run: testing the conservation argument in a coarse-resolution ocean GCM where diffusion and mixing break exact salt conservation. The honest read is (a) — this is a theory paper from a group working at the conceptual-model level, and GCM experiments require different infrastructure and compute. The gamma-kernel analysis also begs for confrontation with observed transit-time distributions from tracer data, which would move the memory-shape parameter from a free knob to an empirical constraint.
