Definitely buffers, child process, event emitter, file system, modules, net (for tcp and low layer), process, streams and some utils (like promisify).

But I'm using Node not only for simple web back end API, but for a lot of IO traffic too (IIoT, Industry 4.0, OPC UA, ...).

You should know the stuff you work with. It does not make a lot of sense if you learn about streams, but never actually use them. Important is how core stuff works: the NodeJS and V8 internals. That kind of knowledge can always make life easier for you, even when using abstract frameworks.

• How does require() work? (hint: put a breakpoint on a require statement and step-in)
• How are modules wrapped for execution? (hint: try executing a JS file with node, which has the content console.log(arguments.callee.toString());)
• What do you have to keep in mind about the V8 work-queue?
• In what way are errors handled (thrown, emitted, returned,...)?
• etc.

Since NodeJS is opensouce, reading the actual source can give you a deep understanding (at least that's what I did)