programming (LP) formulations derived from the multi-commodity network flow problem. In practice, these models take a mining schedule as input and optimise an economic objective (often net present value) through decisions such as stockpiling, processing, blending, and transportation across time and infrastructure constraints. A key modelling challenge is representing stockpiles in a way that is both computationally tractable and operationally realistic. A common approach is to model stockpiles by adding time-based storage arcs, allowing material to remain at a node across time steps. While this is convenient, it effectively turns a stockpile into a warehouse: the optimiser can ‘cherry-pick’ parcels of material regardless of the order in which they were stacked. This behaviour can violate real operational constraints, where reclaim is constrained by how material is physically placed (eg stacking and reclaiming as a queue or a stack). The gap between the modelled and physically achievable reclaim behaviour can reduce plan fidelity and lead to schedules that are difficult to execute. Beyond feasibility, reclaim-order realism can also improve downstream quality forecasting. When a model can cherry-pick parcels, predicted blends (eg forecast Fe on a vessel) may rely on selective reclaim that is not physically achievable. Enforcing FIFO/LIFO ties the planned reclaim sequence to executable stockpile behaviour, making grade predictions more credible. A very simple network flow formulation is provided below for context, omitting the objective function which would generally involve costs/revenues associated with flows on arcs.