It's not necessarily marginal. The EU has 2.5x the population of the US but 20x the languages, address/phone formats, cultural differences/expectations, etc.
When things are changing rapidly and getting new features, localizing and testing that many locales is a large time sink.
I guess it depends on the application, but since we're discussing web app copycats here -- where do you see this monumental time sink?
I'd say l10n and i18n are handled pretty well by pretty much all major web frameworks. Extra work yes, extra testing yes, but unless you've built your app in some insanely obscure way, I can't imagine how localizing it would be more costly than acquiring a copycat competitor, 6 months later?
I believe the obstacles must be more political and operational, not technical.
When things are changing rapidly and getting new features, localizing and testing that many locales is a large time sink.