A highlight should not infer its board position from a translated label

Situation

One visual highlight behaved differently in certain languages and browsers. The rendering logic had become too closely coupled to the displayed label and to an implementation detail that was not supported consistently everywhere. It was a classic UI/data-model leak: text meant for a person had quietly become a machine identifier.

Approach

I bound highlights to stable, language-neutral board and track IDs, and moved the marker implementation back to simpler browser primitives with reliable support. Localized text remained rendered output only. Tutorial boundaries and the keyboard purchase path were also locked down with regression tests.

Outcome

Highlights continued to point at the same game element after a language change, and the browser-specific inconsistency disappeared. The UI carried fewer implicit dependencies on presentation text.