The problem
Take a sealed atmosphere of oxygen and an inert diluent that vents gas outward and is replenished from two finite tanks through valves of limited capacity: the situation of spacecraft cabins, recirculating breathing apparatus and the semi-closed firefighting suit of the earlier paper. A state can satisfy every instantaneous limit and still lack the gas, or the valve capacity, to last the time required. Being within limits now does not say whether the state is viable.
What the paper does
Holding the pressure constant fixes the total rate at which gas must be supplied, so the composition can change only how that total is split between oxygen and diluent, never the total itself. The band of permitted oxygen fractions then acts as an “allocation buffer”: it lets use shift between the two tanks, but it creates no gas.
A simple lemma gives the least of each gas that must be drawn — supply none until the composition reaches the edge of the band, then hold it there — and conservation fills in every split between the two extremes. Intersecting that range with the reserves in the tanks gives exact conditions for viability. Finite valve capacities force mandatory flows that shift the range, and the paper solves that case too.
What it shows
A state is viable for a given time if and only if three conditions hold, for oxygen, for diluent and for total gas, and the longest possible endurance is the smallest of three times: when the total gas runs out, when the oxygen runs short and when the diluent runs short. The paper also derives which starting compositions are viable, the best one to start from, and the extra endurance each added unit of stored gas buys.
In the dimensionless worked example, with reserves split 0.40 and 0.60, total gas is the binding limit at a duration of 71.43. The formulas were checked against discretised linear programs across hundreds of cases, and the largest discrepancy in the allocation endpoints was below one part in a million.
What it does not claim
The paper is explicit that its numerical checks test the formulas against the same model, not against the physical world. The model leaves out carbon dioxide, water vapour, heat, spatial mixing, contaminants, valve and sensor dynamics and physiological limits, and its bounds are symbols rather than medical thresholds; applying it to human endurance or certifying hardware would need a model and validation of its own. The results are open-loop and assume known, constant loads.
Where it sits
The second Galactic Bioware paper. It cites the respiratory-autonomy work as motivation only, and says plainly that none of its formulas transfers to that system as a physiological, hardware or certification claim. Its value is as an exact benchmark against which more realistic models can be checked.