Actually, I was doing some work recently, on some Alexa work. I can't really say much, but I can say that this is definitely a problem in more than the Assistant.
No question, the basic NLU problems have (mostly, as long as your accent is close to the NLU) long been mostly resolved, and things like Now work great in a bunch of languages.
That's a whole different deal from being able to tell that someone said they want to "get down and groove to some blues" or "drown my sorrows in some blues" vs they want to "listen to a blues album" (just a couple of quick examples, look up "NLU example problems" if you really want to read a LOT of crazy language issues that no one ever considers).
Some of these constructs work fine, ported to another language, others work okay, some fail completely.
The Assistant has a lot more skills and such, than Now had, amongst other things.
What if, say I'm in Sudan, and my local NLU manages to determine that I really want to hear a blues album. But nothing is on my device, hmm.... Now in the US, we'd fire off a host of listening and purchase options, probably even over-matching what the user was after. We might even succeed in selling the user something.
In Sudan, well, maybe, there's a music service that offers a limited selection. Due to DRM and other issues though, there might be zilch available, getting back to the "crickets issue". I'm pretty sure they want to avoid these scenarios, based on my experience with other creator/owners of voice-assistant type services, but that's just my experience, a severe under-match has always been a big red flag.
Just throwing up a bunch of search results was considered only marginally better too, than nothing (not exactly "friendly", in accordance with the assistant type model).
The Germany implementation is MUCH easier, plenty of good solid tech "backbone" there, music services with DRM already resolved, and lots of extra disposable capital to drive the whole system, including the development of that language/locale. Sadly, the economics of it really do drive a fair amount of the effort, again, IME.
Not saying that some of these others won't "fill in", eventually, but I'm guessing there's a threshold of sorts, across several areas, that gates the release for a language/locale(s), beyond just a good NLU and intent model.
If you read some of the press releases for Google Assistant, they note that the UK, Germany, Canada, and some others are slated for the very near future. There's one thing in common with all.