Looking into this a bit more, during a single instantiation of DurationFormat, isMatched gets invoked a little over 1 million times.

Some quick optimizations, like memoizing languageMatchInfoLocale.split('-') and manually flattening expandedMatchedRegions instead of using map/reduce, cut the execution time by 30%. More simple tweaks like this can be done, but I expect the real benefit would be finding a cheaper way to implement locale matching.

Thanks for reporting. I'll take a look. Polyfills have traditionally been more locale-data-bound than CPU bound, so for polyfills where the dataset is small we ship them w/ everything.

I’ve seen this issue in older Mono repos.
Clear the cache and re-test localization files using react testing library. This should boost performance.

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

I use this abomination (once, as early as possible) to remove all the locales I don't need:

void function fixDurationFormatPerf() {
    const localesToKeep = new Set<string>(["en", "ja", "fr"]);
    const availableLocales: Set<string> = (Intl.DurationFormat as any).availableLocales ?? new Set();

    for (const locale of availableLocales) {
        if (!localesToKeep.has(locale)) {
            availableLocales.delete(locale);
        }
    }
}();

This function directly mutates the availableLocales internal set, reducing it from ~700 entries to just a handful. Doesn't change the O(n^2) complexity of ResolveLocale, but at least makes n small enough to be usable.