Antibody-drug conjugates have moved from a niche technology to a standard of care across breast, gastric, urothelial, lung and ovarian cancer, yet the three agents used most widely in solid tumours embody incompatible design philosophies. Trastuzumab emtansine maximises linker stability and carries about 3.5 drugs per antibody; trastuzumab deruxtecan carries roughly eight on a tetrapeptide that loses 2.1% of payload over 21 days in plasma; sacituzumab govitecan carries 7.6 on a bond deliberately built to hydrolyse. This review argues that one physical property, whether the released payload can cross a membrane and leave the cell that captured it, organises four literatures that are usually treated separately. The charge state of the released catabolite separates conjugates that kill antigen-negative neighbours from those that do not, and that distinction underwrites the extension of trastuzumab deruxtecan to HER2-low and HER2-ultralow breast cancer, where hazard ratios for progression of 0.50 and 0.63 were obtained without any change to the antibody. Drug-to-antibody ratio is shown not to be independently limiting once conjugate hydrophobicity is controlled. Resistance is moving away from the antigen and towards the payload: HER2 is lost or reduced in roughly two-thirds of tumours rebiopsied after trastuzumab deruxtecan, but efflux transporter upregulation, SLC46A3 loss, TOP1 mutation and SLFN11 loss now account for resistance in models where antigen expression is intact, which argues for switching payload class rather than target at progression. Interstitial lung disease is examined as the clinical expression of the same permeability. Adjudicated drug-related pneumonitis affected 15.4% of 1150 patients across nine trastuzumab deruxtecan studies, with 2.2% fatal and a median onset of 5.4 months, and the available mechanistic evidence points to target-independent uptake by alveolar macrophages rather than on-target HER2 binding. The consensus management algorithm is set out in full, together with the finding that no dataset yet demonstrates that its corticosteroid regimens alter outcome.