Zotfile allows automatic renaming of attachments, which includes moving the PDF file to a different folder. This means my zotero account still has plenty of free storage, while the space-consuming PDFs are stored in my university's effectively limitless cloud drive.
I use Recoll to fuzzy-search the text of the PDFs, or just pdf2text piped through grep for simpler searching.
Recoll is great, it's here: https://www.lesbonscomptes.com/recoll/
Like many others, I write docs in markdown and use pandoc and citeproc to render to Word, then libreoffice to convert to PDF since it reliably retains the formatting from a lightly-styled Word template. With citation style sheets I can be reasonably confident that citations and references will automatically take care of themselves - being tweaks for edge cases.
I use pp to pre-process the markdown, mainly to allow "includes" so I can write chapters/sections as separate files then combine them. All using a horrible render.sh script that I copy into every new project and tweak as required :)