Every UI string is rendered with the real font metrics for each language, not an estimated character count. German and Finnish compound words routinely run 30–40% longer than the English source; Arabic often runs shorter in character count but each glyph is visually wider; Japanese and Chinese pack the same meaning into far fewer, wider glyphs. In Fixed-width mode the button/label keeps its English-sized box and clips overflow text with an ellipsis — a real, common localization bug. Flexible mode lets the box grow (or wrap) to fit the actual rendered width, which is the correct fix.
The cart toast separately tests ICU plural rules. English has two grammatical categories (one/other). Russian and Ukrainian have three (one/few/many) selected by the last one or two digits of the count. Arabic has six (zero/one/two/few/many/other). A codebase that naively reuses the English n===1 ? singular : plural rule for every locale silently produces the wrong grammatical form whenever the target language's rule disagrees with English at that count.
ru/uk: n%10==1 && n%100!=11 -> one
n%10 in 2..4 && n%100 not in 12..14 -> few
otherwise -> many
ar: n==0 -> zero | n==1 -> one | n==2 -> two
n%100 in 3..10 -> few
n%100 in 11..99 -> many
otherwise -> other
en/de/fi: n==1 -> one, otherwise -> other (2 forms only)