All the below, plus
Every time you take a shot a certain amount of read noise is generated. This read noise depends on the camera and the gain being used. For most DSLRs this can be quite considerable compared to CMOS astrocams. Some websites have this listed, or sharpcap has a routine to calculate this and more.
So, for a certain total exposure, the total read noise will be less if you take longer (and therefore fewer) sub-exposures as the read noise adds up. But one of the downsides is that longer exposures will lead to saturation of bright stuff (stars).
What many try to do is have a long enough exposure so that the background level 'swamps' the read noise. One such criterion is background noise > 3x read-noise. And note that a shorter exposure time is required to swamp the read noise in worse light pollution. Or you can look at as the overall signal-to-noise ratio - the improvement in this reduces as you increase the sub-exposure length.
And it also depends on the speed of your scope, etc etc. I made some spreadsheets to calculate it all (basically stole the info from Shiraz, CloudyNights topic 536809). Its fun to do. But I'd ignore everything I've said and use the longest sub-exposure where you only get saturation of the larger stars.
Chris
