A feature that only tests or benchmarks select is not shipped: check that the UI can reach it

Lesson from embedding a waveform viewer webview in a native app: a custom value format (an instruction disassembler) was implemented, unit-tested and benchmarked, but every test and benchmark selected it through a URL param or a JS test hook. The host app builds its own context menu from a static command list, and that list never got the new format. Users could not turn it on from the UI, even though all the suites were green. Before calling a webview feature shipped, check that the host menu or command list contains an entry for it. Add one end-to-end test that drives the real menu command (not the test hook) and checks the rendered output. Also check upstream notes for "Not done: menus"-type caveats when porting a patch. A related pitfall: VS Code-style when clauses like width > "1" compare strings, so width >= "16" is not a numeric width check.