The way my data works isn't easily caught in terms like chapters, so to be a bit clearer I will redefine some things.
I will have 66 chapters and about 1200 sections, each section containing several lines. (estimated no. of pages is 3000-4000)

I will go through your questions 1 by 1

• How are your headings/titles to be identified from those data?
every line in the csv starts with the chapter, then a section id (resets every chapter), a line id (resets every section) and the content itself (with tab as separator).

• What determines where the content for one heading ends and the next one begins?
Every line contains what chapter it is part of, so when that changes, a new heading needs to start.

• Can the content relating to a given heading/title span more than one page?
Section usually span several pages, lines (in the csv) never do.

• Does each heading/title begin a new page or can it start part-way down a page?
To reduce paper use and thus price, I mean to reduce as much white-space as possible, so it is fine if a chapter starts part-way through a page and with the sections it is even preferred that they do.

• Can two or more headings/titles and their content occur on the same page?
If a chapter happens to end, there would be content of two chapters on 1 page.

• What format is the output to take (e.g. normal print, booklet print)?
I mean to send it to a publish-on-demand business. I don't think I understand your question in more detail. Ultimately I will need to send them a pdf.

• Does it matter whether the data is stored in the application in the same format as the final printout (i.e. does it matter whether you can see all columns side-by-side in the electronic file)?
No. As long as I can check that in the final printed book it will be side-by-side (also across the 2 pages), the visibility in the 'original' electronic file doesn't matter.

Hope you have enough answer here.