I absolutely agree.
I've actually had this thought lingering in the back of my head as I'm in the midst of making the transition from primarily MS desktop/embedded C++/C# guy to open source stack web app guy.
A few years back, in a review my supervisor put in a recommendation to go into management. I asked if I would be able to code anymore. Nope. I felt that was some subtle death trap - to lose my skills through non-use, and turned it down, 20% pay bump be damned.
So now this transition, and began to think of it in the same light. But it doesn't work. It's been 13 years since I graduated, and I'm probably better at most CS'esque interview questions than anytime since then. Why? Because of what I did that led me to making this transition, essentially becoming unhappy in my job and looking for anwers, which started with a book called "The Passionate Programmer", which led to maybe 20-30 other books, completing some classics (such as K&R C), some incomplete (SICP, only halfway thru chapter 2), but in general exploring my craft in a way that my job did not require me to. I'm a much better programmer because of it, language be damned.
So here's the question I have:
Who is more likely to be aware of things such as CLRS, SICP, Hacker News, Stack Overflow? Who is more likely to understand (or even be aware of) languages like Clojure, Haskell, Prolog, Python, Ruby, etc? Who is more likely to give you more lucid answers to problems with big data, having even touched a NoSQL database and pontificate on the advantages/drawbacks of that vs. relational? Your typical enterprise dev, or your typical open source web dev? After talking about all this stuff to a team of 30+ engineers over the past few years, I'd have to put my $$ on and double down firmly on the latter.
Of course, this is only an example of one engineer's experience. But I think there's a more meta issue at play here... it's more of the 501s vs. the non-501s, the latter being ones who well, just are more aware of what's actually going on in their respective industry. If the industry as a whole makes a tectonic shift toward algorithmically more difficult problems (ie. machine learning), I trust the open-source web devs will be on it and synthesize it faster than your typical enterprise guy will even be aware there is a need.