Thought Numerical Recipes was used mostly in courses for students to learn about numerical methods rather using those pre-made routines for building large programs.
Most people using NR aren't building large programs.
Think about a student who learned from NR, then in grad school needs a method, and in the spirit of "I'm smart, I can do this myself", copies the method out of the book.
In larger projects, someone is more likely to be aware of alternatives, and the issues related to using source covered under the NR copyright.
Back in ... 1991? I did use the C version of the Simplex Method almost straight from the book. I needed a concise linear programming optimization module that would be called in the middle of a not-really-concise production programming system handling (automotive) engines and gearboxes production for a couple of FIAT plants.
Basically a short-period scheduler to decide what to produce (slave to an immense system running on IBM mainframes; COBOL, again).
So I quickly put together the C version (with minimal modifications) and it chugged along happily on a VAX for at least a decade.
To explain my frame of mine, I was thinking how it was used in 2007 or later.
In the early 1990s, there wouldn't have been much off-the-shelf/free&open-source options. The GNU Scientific Library didn't start until 1996, for example (according to Wikipedia).
Think about a student who learned from NR, then in grad school needs a method, and in the spirit of "I'm smart, I can do this myself", copies the method out of the book.
In larger projects, someone is more likely to be aware of alternatives, and the issues related to using source covered under the NR copyright.