
Elaine Rich's textbook "Automata, Computability and Complexity: Theory & Applications" https://www.cs.utexas.edu/~ear/cs341/automatabook/index.html
Useful because it covers non-deterministic Turing Machines complexity classes like BPP,ZPP,RP,etc.. that you would run into Monte Carlo methods etc. As the only Non-deterministic Turing machine most people have been introduced to is the maximally lucky guesser from NP's definition it can help with practical solutions.
Unfortunately it doesn't have bounded-error quantum polynomial time (BQP) which would help set expectations for quantum supremacy.
Furthermore, most CS grads I know would benefit from some statistics knowledge. I've even found physics, linear algebra, and biology (specifically genetics) very useful on the job.
I see academic subjects the same way many people see programming languages -- they're tools. If all you know is programming, everything starts to look like a programming problem. There are other kinds of problems, too, and even as a professional programmer I often stumble across problems where (basic) knowledge in another field has helped immensely.