In revising a 1953 manuscript in which the author (a non-native speaker, a professor, quite proficient in English yet with a few lingering idiosyncrasies) and the original editor (OUP, New York branch, still working under some British English conventions in house style) made often inscrutable choices in the use or non-use of commas, I'm authorized to make alterations in this area, but I'm first looking to develop some operational principles, a rationale for when to alter. We know commas can be tricky because they're sometimes optional. I'd prefer to error on the side of minimal intervention, making changes not when the comma or lack of one is a mere felicity or stylistic matter, but where their use genuinely facilitates ease of reading and comprehension.
I'm looking for signposts, guidance, a good reference as to when this is a matter of discretion and when the rules are hard.
Reviewing the 400-page MS, it's sometimes obvious when to intervene. Drop it into software like Quill Bot or Grammarly for a quick sense, and every page, if not every paragraph, has numerous comma recommendations.
How would others approach this? Please refer to any CMOS sections that focus on the discretionary aspect.