I did it in 10 days. Painful. But possible.
Chapter 1 intro/lit review summarized the organization and cribbed some applications and overall intro stuff from a school paper I had done. Chapter 2 methods was synthesis of methods from 4 papers. Then 3, 4, 5, 6, 7, 8 were just the specific papers (cutting the methods section and intro and light editing for consistency). 4 of the 6 papers were already published. The other 2 were close to publishable (work done, wrote it up).
I did not bother with a general conclusion chapter (3-8 chapters had conclusions by chapter within them). Avoided the LaTex rathole and did it all in Word (significant amount done from published papers already). Didn't do a large Word file as Word tended (back then and probably still now) to break down for book length projects...and dealing with section breaks and the like is a pain (easy to mess up). Did each chapter as a separate Word file.
Endnotes were by chapter and at the end of whole thesis. Figures not embedded, but at the end of each chapter, just like a paper submission (even copied one diagram in on a transparency, totally old skool).
Did a chapter a day and had 2 days at the end to compile and print (used Kinkos for the print job). I typed at night, all night for isolation and concentration (this was before PCs at home were common, so in the lab). Would go for a run and eat breakfast and sleep in the day. Very disciplined.
Set all the pagination of the different files at the very end manually to make it all work (once edits done). Then printed all the different files and compiled them all (then went to Kinkos). Even did the TOC/TOT/TOFs manually (was easier to just kludge it).
Obviously follow the rules for margins and double space and all that.
I offered the advisor to see each chapter as I finished it, but he didn't give any input and huffed a little about how most students did their thesis with more warning. (But really the papers were all mine 100% in content anyways.) I just told the old man that I was treating him with the same rules as the rest of the committee (10 days to read and then a defense). If it wasn't good enough, he could fail me. He wasn't going to do that since I was a little golden boy with national award and first to graduate and fancy job (and really I had aced the Ph.D. program, written lots of papers, so why fuss about the thesis drill).
It ended up being fine. And was way faster to just hack it together like I did than people who told me I should spend a few months learning LaTex at the last minute or stuff like that. Just saw so many people struggling with huge documents but my hacking it into little Ricci flow segment approach worked great for being under the gun.