Ask HN: How to best acquire theoretical computer science knowledge?

Theoretical computer science is probably not what you mean. Theoretical computer science are things like complexity theory (up to arithmetical hierarchy and randomized and approximation algorithms), theory of recursion (up to e.g. Gödel's theorems), (various kinds of) logic, algorithms and datastructures (full of hardcore probability and statistics), etc.

Subjects like that give you a firm grounding in the area of computer science but you could study them for ages especially on your own. These subjects are also not very practical in terms of everyday engineering and on top of that before you even get to that level, a top school would require you to study a bunch of mathematics to gain a certain level of rigor of your thinking. My guess is that this is hard to do on your own.

On the other hand, since you have a related degree already and since you know programming, I would heartily recommend you to study on your own combining two approaches: bottom-up to learn the basics and top-down to immediately start increasing your market value and to start opening doors and to keep yourself motivated to go forward with the more tedious study of the basics.

E.g. say you are interested in distributed systems - you can start reading up on them (top-down) while digressing here and there to learn some of the basics.

Make use of online resources like Coursera for the necessary basics (bottom-up): at least some maths heavy on proofs [1], at least some complexity theory, at least some graph theory, ....

I think a purely bottom-up approach (going to a university) could be a waste of time and money in your case. But you need to emulate it a bit because a purely top-down approach could be too superficial and wouldn't teach you some of the more rigorous thinking you might need.

Whatever you do and learn try to do it in-depth. Superficial knowledge of many things will not benefit you much in the long term while in-depth study of many computer science fields will likely lead you through a series of small enlightenments.

[1] My professor of linear algebra used to insist that we as computer scientists have to know all the proofs of the theorems we were learning - in contrast to mathematicians learning the same subject - simply because we have to know how things are done to the last detail and we have to develop that kind of thinking.

Any reason to go for Sedgewick and Wayne over CLRS? I haven't read the first one so just curious.

I concede it's not the best approach if someone has a specific goal like, as you suggest, becoming a "data scientist", a "compiler developer", or a quant, etc.

In my case (and what I felt was the case for the OP) a general, broad level of knowledge is paramount. I then study deeper on an as-needed basis off the back of my broad but shallow knowledge. Being able to tie together the basics of many unrelated areas really helps in my work, but is surely not for everyone. I already have my dream job for life so my goal is more on becoming well rounded rather than getting a specific job.

From a career POV, should the OP specifically focus on a specific target rather than just aim to get a general CS understanding? Quite possibly! But that's a different discussion/recommendation that I think you could kick off in your own response to OP. So I'm voting you up not because I agree with your criticism, but because perhaps the OP does need to think more solidly about the end goal.

I second Sipser's book. Probably the best textbook I used in university. You can find international editions for a very low price.

I am suggesting an approach rather than a source. Sources, as someone already said it, are far from rare.

I would say pick up a single system, say database (relational or document) or routing, OS, DNS, Web server, A compiler, Filesystem, NFS, etc. and study the hell out of it.

Once you know how to understand a particular system in depth, everything else will start falling in place.