Stacking individual RAW files vs stacking integrated FITS files

I would definitely stack all the data from each telescope as master lights, then take those master lights and run those as a mosaic if there are big differences in FoV and overlap, otherwise as a normal integration with same camera and optics unchecked. On the normalization tab make sure you have it set to advanced, then on the integration tab uncheck MBB and set local normalization correction to 4th degree and 5 iterations at the least. That gives me good results when stacking multiple telescopes. It is also much faster because you aren't trying to register and normalize multiple FoVs against each other as individual raw files.

So the steps are:
Stack the lights from each telescope as master lights.
Put those master lights in as lights after you have all your data together (if combining OSC and NB data split the OSC into channels and then stack those as master lights)
On tab 3 set it to 2000 stars (trust me on this one)
On tab 4 uncheck same camera and optics, and if you are doing a mosaic set it to that with a 1-10, 5-15, or 1-15
On tab 5 set mode to advanced
On tab 6 uncheck MBB if you have it checked (I normally leave it checked and at either 10 or 15 percent), then set local normilization correction to 4th degree 5 iterations. I normally leave it at 2nd degree 3 iterations for most of my stuff.
Then hit integrate.

As a note if you have large variance between the pixel count of your cameras you may want to either decrease the size of some of the large pixel count ones or increase the size of the small pixel count ones. I learned that with combining data from an FLI PL16083 and QHY600M Pro.