I just did this recently with nearly 100 chapters in individual doc/rtf files. This was my basic procedure:
1. I first converted the files to txt with LibreOffice command-line.
2. I wrote a Python script to parse the txt files, figuring out section headings, cross-references and verses. It also cleaned out some junk and converted some code-points to update the encoding. Some files still had \v markers (originally exported from Paratext) and others didn't so I just had to go by the numbers in the text. The Python script output SFM
3. I iterated back and forth between updating/refining the code and making some manual edits to the txt files to clean things up until I had it in pretty good shape.
4. The last step was to concatenate the chapters into one SFM file, and then I copied and pasted that book by book into the Standard view in Paratext. (I had been recommended this years ago as the safest way to paste data, as PT will do some checking and tidying up of the formatting when you do this.)
5. If there were major issues (e.g. \s markers were missing \p after them so half the chapter formatted as section heading) then I'd go back to step 3. Otherwise I'd fix up any small issues manually within PT itself.
I've been encouraged to put my code/procedure online and hope to do that soon. If it might be useful to you, I can get onto that sooner than later...