I strongly(!) recommend that you do not try to “switch during a single render.” That is simply “not how real movies are done!”
Here’s what I do, and it works:
(1) Create a “base scene” in a render-file, then “link” additional scenes to it in the same file. Each one of these clearly-named scenes represents a single “camera shot,” along with the file-output parameters (filename and directory) that are specific to it.
(2) Initially using a “quick and dirty” render option (Workbench), I “shoot film.” Lots of it. “With reckless abandon,” if necessary. Each frame is “stamped” with relevant information. (The shots, sets, dummy objects, and everything are carefully set to physical scale. File-linking and so forth are also set up as they eventually will be, even if most of it hasn’t been “designed” yet.)
(3) Next, go to “any video editor” – Blender’s own, open-source, or commercial – to begin to “cut the film together.” This approach has been referred to as: “Edit, then(!) shoot.” You can get amazingly far. You can afford to experiment. “The power of human imagination” is a wonderful thing. 
(4) Now, shot-by-shot, begin to replace the stand-in footage with “real stuff.” (Make new copies of the blend-files and keep the stand-ins for reference. They should always match exactly.)
You’ll actually spend a lot of time in “stage 3,” but the critical observation that it is cheap. You can explore many different options for the pacing of the general show, and you can actually “very fine-tune” individual shots. All with “stand-ins and fast renders.”
This lets you, when it comes to “actual rendering,” be very precise as to exactly what you need to render … and to model, texture, and so on. If you already know in advance that a particular thing won’t appear, you don’t waste time building it. (Just like the “paper-thin towns” of all those Spaghetti Westerns …)