In brief, I used Scrivener for writing the thesis text itself. I had already written quite a bit in Markdown using various text editors, including DT. I used DT for notes, especially reference notes. In terms of scripting, I mostly used it for pulling reference information across from reference managers (Bookends, Zotero). I also indexed all my reference PDFs in DT. I used DT partly to organise my information and thoughts, including idea notes and topic notes with references to other notes (sort-of Roam-like or Zettelkasten), and also to search — I found the search incredibly useful especially at the end when I needed to find hard-to-recall specifics to fill in footnotes or provide detailed justification of a point in my argument. At one stage, I worked hard to build a map between DT notes and other notes in TinderBox, but I never really made much progress with this.
I did all my formatting in Markdown/LaTeX. I have used Nisus Writer Pro extensively in the past, and would use that (perhaps with Adobe InDesign for final layout), but the problem in my field (Biblical Studies) is that the main style guide for formatting citations is almost impossible to adhere too outside of BibLaTeX and the biblatex-sbl package. Hence I decided to stick with LaTeX for layout/formatting.
I do have hopes that as Zotero, and hence CSL-based processors and in particular pandoc-citeproc, make progress towards more complex style implementation, I will one day abandon LaTeX altogether, so my main investment is in pandoc-flavoured Markdown. Scrivener compiles easily to Markdown, DT works well with Markdown, and I can always edit in a text editor, so it feels like a plausible long-term investment.
The biggest change I am planning in the immediate future is around citations. As you can see from the scripts I’ve posted in the past, I always struggled with the gap between my reference manager (Zotero for a long time, and Bookends) and DT. I only really need a reference manager to generate the citation entries, at the moment in BibLaTeX format, and perhaps also in JSON for CSL formatting. I would much rather work with my PDFs, EPUBs, random docxs, and all my literature notes in one app, namely DT. It is not too hard with custom fields to add the reference information (a bit clunky, but then Bookends and Zotero are both clunky in different ways too). Then it’s just a matter of scripting the generation of the appropriate BibLaTeX code fragments for each reference. Funnily enough I have a fair idea of that code because in both Bookends and Zotero, the demands of the SBL style meant that I had to customise the output pretty heavily anyway.
Once I’ve done that the reference manager piece goes away, and I’ll be working almost exclusively in two apps: DT and Scrivener, with pandoc and LaTeX behind the scenes doing the formatting.
Hope this gives some idea of what I’ve done.