A viable solution for Python concurrency

Well, it depends on how you look at it. If you're, say, an application developer, it's easy enough to say "that seems like a bad idea, let's not do it." And in that case (i.e. the case where there's no actual mutation of shared state), the performance penalty seems pretty minor to me (if I'm understanding the process you describe correctly, and in comparison to the existing CPython overhead, which is quite large to start with). On the other hand, if you're a language developer, you don't really have the luxury of saying "this seems like a bad idea, let's not support it." Somebody may have already deployed a real-world application which depends on this behavior, so you can't unilaterally break it unless you want another backcompat flag day a la Python 3. Nobody wants to do that again.