Here's my workflow in a nutshell: I use the Zotero browser extension to add items to Zotero, the extension takes care to automatically attach a snapshot of the web page or paper to the newly created entry in Zotero. For academic papers or blog posts, I'll try and extract additional web links (e.g. Github repos) and add those to the entry as well. Within Zotero I have a nested structure of collections/folders to roughly structure the entries by type (papers, blog posts, reports, books, ...).

I make liberal use of tags to classify the content and keep track of my reading status (Reading List and Read), for which I have saved searches in Zotoro, so I can easily find something to read, whenever I have a bit of time.

For academic papers that I've read in-depth, I'll extract the most relevant references and add them to the library as well and cross-reference them using the "Related" feature. This feature is what I was mentioning in my original post - it only allows for a single generic "related" association between entries. If I could customize the type of relation, I could model almost arbitrary knowledge graphs here.