Linux

It can recognize tracks based on a profile of the audio content in the file, so even if they're not labeled properly it can (usually) identify them. The biggest problem I've had with using it is that many artists have re-released or rerecorded songs, or released the same album in different years in different countries but maybe with one track missing or with the tracks in a different order, and sometimes Picard will pick an album release that you know is incorrect. It will reorganize the files, and there are lots of options for controlling how it does that (e.g. file renaming and tagging, output to the same folder or a different one, automation of bulk processing). But you may have to guide it a little, especially if you have many albums from the same artist because then you'll have overlaps.

I've found the best approach to handling this is to work on one folder/album at a time and make sure it picks the correct release year for metadata reference.

This is the right answer. I specifically don't use beets because it does this, and (1) I have things organized how I like, and (2) I've changed music servers a half dozen times in the past decade and am leery of an app that has to "own" the library.

But the reasons I don't use it are exactly the reasons you're looking for, and it sounds like a perfect fit for you.

Ah interesting point, I've only done this when importing new songs, not to an entire library. Usually drag my new files from a file browser into a new playlist then I select then all and right click, then hit "move to library" which brings up a dialog that allows you to define your folder structure and naming conventions