The oxygen balance of an oxygen minimum zone (OMZ) reflects competition between respiratory consumption of sinking organic matter and physical resupply. How much organic matter is respired has received far more attention than where it is respired, yet that depth determines whether the demand acts on the OMZ core or the oxygenated water above it. This study quantifies the control with a one-dimensional offline plankton–oxygen model at five Bay of Bengal stations spanning a sevenfold range of core oxygen, perturbing the detrital sinking speed and remineralisation rate with every budget closed from recorded increments. Moving from the weak-transfer to the baseline configuration displaces respiratory demand downward and increases it eightfold to thirteenfold within a baseline-defined core layer. The resulting loss of oxygen contrast across the core upper interface reduces the downward diffusive supply by -37 to -9 %, with the larger responses in the three oxygen-depleted columns. That amplitude is set by prescribed closures rather than by the particle physics, spanning -14 to -48 % across perturbations of vertical diffusivity, restoring timescale, oxygen half-saturation and diagnostic layer width. It survives removing oxygen restoring from the upper 50 m but collapses, and at one station reverses, when the upper 100 m is freed: the response requires that the water above the core not deoxygenate with it. At one station, configurations sharing a constant ratio of the two parameters agree to about 1 % across the tested triplet, so oxygen alone separates them poorly over that range. Independent biogeochemical Argo float profiles, used for evaluation only, place the modelled oxycline in the same 25 m bin as the observed one and reproduce the station ordering of core oxygen, but do not constrain the interface gradient, which varies by more than a factor of two between defensible processing choices. The mechanism therefore remains untested against observations.