Background. Stretch-dominated lattices such as the octet truss are preferred over bending-dominated architectures such as the body-centred cubic cell because their stiffness scales linearly with relative density rather than quadratically. The comparison is usually made for the ideal geometry. Manufactured lattices contain strut thickness variation, waviness and missing members, and the two topology classes do not tolerate those defects equally: a stretch-dominated lattice carries load through determinate paths, while a bending-dominated one is highly redundant.

Objective. To quantify the stiffness and strength advantage of stretch-dominated topology across relative density, to identify where strut buckling rather than yielding governs collapse, and to determine how much strut loss the advantage survives.

Methods. Analytical Gibson–Ashby scaling relations were evaluated for both topology classes in an aluminium alloy of 70 GPa modulus and 250 MPa strut yield stress, across relative densities from 0.02 to 0.35. Collapse strength for the stretch-dominated lattice was taken as the lesser of strut yielding and elastic strut buckling, with the buckling coefficient calibrated so that the two coincide near a relative density of 0.05. Imperfection sensitivity was evaluated by Monte Carlo over populations of 4,000 struts with log-normal thickness variation and a randomly removed fraction, treating the stretch-dominated response as governed by the weakest members of a critical load path and the bending-dominated response as governed by the surviving population mean.

Results. The stiffness advantage of stretch-dominated topology fell from 11.3-fold at a relative density of 0.03 to 1.1-fold at 0.30, so the architectural choice matters most precisely where lattices are lightest. Strut buckling governed stretch-dominated collapse below a relative density of 0.050, reducing strength at 0.03 from 2.46 MPa under a yielding assumption to 1.45 MPa, a 41 % overestimate if buckling is ignored. Imperfection sensitivity was markedly asymmetric: with 15 % strut thickness variation and 5 % of struts missing, the knockdown factor was 0.606 for the stretch-dominated lattice and 0.975 for the bending-dominated one. The strength advantage nevertheless persisted over a wide range, falling from 2.42 to 1.96 at a relative density of 0.10 as strut loss rose from zero to 10 %, and reversing only at 36 % strut loss. At a relative density of 0.25 the reversal occurred at 20 %.

Conclusions. Stretch-dominated topology remains the correct default, but its margin is smaller than ideal-geometry comparisons imply and shrinks with both increasing density and increasing defect content. Two design rules follow: check strut buckling explicitly below a relative density of about 0.05, and prefer bending-dominated architecture where process control is poor or where damage tolerance rather than initial stiffness governs.