I don't think there is a clear path, or a collection of clear paths. If you're interested, read stuff and try stuff. With hard work and enough luck you might be a "wizard" some day. It would be silly to make promises.

For whatever it's worth, I have a PhD in compilers, and I work as a compiler developer. I have done nothing listed in the article except learn about regular expressions.

Also, I disagree with your disagreement with the last part of the article.

I actually want to pursue a career around engineering compilers. There's a company I'm planning on applying to next year: at the moment I'm studying a part-time MSc in computer science while working full-time in operations for a development company.

What resources have you found helpful for your Ph.D.?

In your opinion which journals/articles are classic/must be read in this field?

Is there any topics (or books covering these topics) that are a "must know"?

This is a very broad question, since compilation is a very broad field. You should get a broad overview of the whole area through one of the standard compiler textbooks. I think the only one I have read cover to cover was the Dragon Book. It was one of the older editions that are no longer up to date; I'm sure the newer ones are fine. I'm also hearing very good things about Appel's "Modern Compiler Implementation". There are others as well. Read one, and follow the literature references for whatever areas interest you most.

If there is one "cross-cutting concern" in compilers, it's the importance of program representations. The correct representation will allow you to do things that you wouldn't be able to do otherwise, or only at much higher cost. So some more concrete things to look into are SSA form and the Sea of Nodes representation (for the latter, Click: "Global Code Motion/Global Value Numbering", https://courses.cs.washington.edu/courses/cse501/04wi/papers...). Some general graph algorithm stuff (depth-first search, cycle detection, dominance) is useful.

One surprisingly commonplace, simple, and very useful thing to know is the Union-Find data structure (https://en.wikipedia.org/wiki/Disjoint-set_data_structure). I've used it in various settings in compilation. Once I was doing something in GCC and needed a union-find implementation; poking around, I found three or four of them. None were reusable, so I added one more at the place where I needed it :-(

As for journals/articles, much of the seminal compiler work appeared in the proceedings of PLDI, but that doesn't mean that it makes sense to methodically go through 30-odd years of historical papers. ACM Computing Surveys (https://dl.acm.org/journal/csur) can be quite good, if they are not too old. If you are looking at a specific area and see a reference to a survey paper on that area, definitely follow it. But all this doesn't mean that you should only focus on certain conferences or journals. If you want to dig PhD-level deep, it's very much about following references in general.

Good luck! Let me know if you have further questions.