In the C family of languages all the notable compilers either never used a parser generator and instead simple recursive descent, or moved from using one. It's easier to write, debug, and extend an RD (or table-driver precedence-climbing) parser than work with the (large amounts of) generated code tools like Bison will create.
In general, I don't think so. It's one thing if you parse a well understood grammar like C. But many people who use recursive descent from the start for a new language or DSL just don't bother about the grammar, and when someone takes a look it is suddenly LL(4) and relies on whatever the hand-written parser does.
So what? I personally am a fan of recursive descent parsing for most programming languages, and if it works with recursive descent, why would it be troubling that it's actually an LL(4) grammar?
Now it would only be troubling if the grammar itself is intended to be widely usable across different languages, say JSON or XML. A parser generator really shines because it can generate code to parse this grammar in multiple languages. For programming languages intended to be written by humans, I value things like good error reporting far more.
