I have seen quite a few images lately with people saying things like "This is 48 hours of data" or even greater integration times than that.
But that still left me wondering, how much is too much? At what point will I experience diminishing returns as I add more data and see less and less improvement?
The Signal to Noise Ratio (SNR) improves by the square root of the number of frames. Knowing this, it was easy to plot these points and create the charts below.
SIGNAL TO NOISE RATIO FACTOR
The SNR Factor chart below shows the improvement to the Signal/Noise Ratio as more frames are added, increasing 50 frames at a time.

By adding a linear trendline, we can estimate a point of diminishing returns at the intercept. However, the trendline crosses twice (at 700 and again at 3800 frames). This tells me two things- there will never be a point of negative returns where adding more frames decreases quality (shocker), and there is no single point of diminishing returns but instead a range. This then becomes of exercise in Managing Levels of Effort- At what point is "Good enough" good enough for you? Each set of images will take more effort to collect and process for less return. You will have to weigh your effort to the gain you receive for the images.
Marginal Rate of Change
The Marginal Rate of Change chart shows the percent of improvement to the SNRF that each new set of 50 light frames will add. We can calculate this by taking the first derivative of the SNRF. 50 Frames are 29% better than 10 frames. 100 frames are 22% better than 50 frames, and so on. Even at 10,000 frames, you will see a 2% reduction in noise for every additional 50 frames you add. 
Based on this, I can confidently say that adding more frames will always improve the quality of your image, but not always in a meaningful way. The point at which good enough is good enough is a matter of how much effort it takes you to acquire the frames and the quality you desire.
I travel to take pictures rather than take photos from my backyard, so I have a considerably high level of effort for each set. At 700 frames, each additional set of 50 frames will improve your image by less than 10%. At 3800 frames you'll only get roughly a 4% decrease in noise for every additional set, and by 10,000 frames you will be seeing a 2% increase per set. Where you draw the line is up to you.
Final Thoughts
This principle will apply to any set of light frames no matter what the exposure time, so long as all the frames have the same exposure settings. The length of each exposure will affect how much noise is collected in each frame, but the number of frames stacked will affect the SNR of the final stacked imaged and that's what I was concerned with here.
Don't waste a clear night taking 700 out of focus light frames. Get a bahtinov mask and get tack-sharp stars every time with ease.