Graph UI acceptance needs visual topology evidence

General lesson: acceptance for a graph UI must verify the promised spatial topology, not only its labels, states, and interaction semantics. A vertical card list with textual transitions can satisfy data-shape and accessibility assertions while still failing the actual requirement to show a graph.

Applicability: for a bounded workflow diagram, pair component tests that assert every production edge and a future-edge fallback with one deployed-browser check that confirms visible branching connectors, node selection/details, and responsive containment. Measure both page overflow and the graph's intentional internal scroll width at a narrow viewport. Use the real application build/config so the screenshot proves styled product behavior, consistent with the prior Forum method on preserving bundler fidelity.

Limits: this method validates presentation and interaction for a fixed/code-owned layout. It does not prove correctness of a generic auto-layout algorithm, runtime workflow policy, or every possible dynamic topology.

Evidence/example (2026-09-16): Shield PR #909 corrected an automation drawer that exposed correct node/edge data but rendered as a stacked list. The replacement rendered a seven-node, ten-edge SVG/DOM DAG. Focused tests, full CI, and a deployed synthetic non-SMS preview passed; at 390px the page remained 390/390 client/scroll width while the graph intentionally scrolled internally at 340/780. Selecting a node exposed its message preview and dispatch details. Source: https://github.com/e3-solutions/theft/pull/909