Just to clarify, when I say "theory of languages" I mean a course that compares different ways of expressing constructs as well as different constructs. For instance a language may implement a variant of for...loops that allows each iteration to be done in parallel, not just in sequence. That will typically require you to learn lambda calculus, at least a bit, but this makes you a better developer anyway.
A TOL course will typically compare and contrast different languages so although it's possible to do loops in Lisp or it's bastard child SQL, that's rarely elegant or efficient.
This may intersect with learning about compilers an activity that has declined from the core of CS courses to the periphery, partly because CS has grown and partly because the quality of CS undergrads has declined and many find the idea of syntax being anything other than something you learn from your betters as uncomfortable.
A compiler course is of use to pretty much exactly one type of CS grad considering some form of quant work, the very high end number crunncher, numerical analysis and including supercomputing & GPU type stuff. If you want to write stupidly efficient code you have to know how it will be translated and optimised.
Note that I say "stupidly" efficient, in nearly all the code you will ever write +/- 20% doesn't matter at all, and >80% of you code could run 5 times slower without anyone noticing or caring, so it's not worth your expensive time to save CPU. For instance last week I upgraded my main PC from 6 2.8 GHz cores to 8 * 3.5 GHz, making it about 30% faster for about $150, at a bank you will cost (not earn) that much per hour. My change makes all CPU bound code run faster with very low risk that some will stop working.
I have spent whole days trying to make code go faster with no result, in face in some cases, I've noticed bugs or edge cases that caused me to add code that slowed it down, making my apparent productivity negative.
So if you see your future as writing ultra sharp code, do the compiler options, else do some maths.