I wish that that someone did some tests between old and new software, I bet uptime of old software/hardware would be higher than our current tech stack.
So we're far from unhackable, never-crashing software, but then cars, too are still relatively dangerous, just much less so than they used to be.
(Now one interesting thing that I wonder about, though, with all those thank yous going out to the government, Ralph Nader and other deities in sibling comments, is just how reliable this crash test is; meaning, what happens if you try other angles, which are not in the standard test.)
BTW, Java does have implementation-defined or unspecified behavior (for instance, using the same objects from different threads without proper coordination), but there's much less of it and it's easier to avoid.
That said, AFAIK language-level invariants in Java (and Python, etc.) cannot be violated by the lack of thread safety; meaning, you can still inspect every live object and see its state. Program-level invariants might go to heck, for sure. So what I should have said is more along the lines of "memory safety" than "lack of undefined behavior"; I'm not sure what the shortest and most precise way to put it is, though. (It does not quite end with memory safety but it doesn't quite get to lack of undefined behavior, either.)
A lot of work has been put into tools and our general ability to write better code (although whether we do so or not is a different issue).
Would you like to write a web-app in C?