Caveat: my experience is with reflection seismic surveys, but I think many of the principles are similar to MASW.
As you guessed: it's an optimization problem. There are lots of factors at play. It's up to the geophysicist to balance the various needs of the survey:
- You need to image the target with useful accuracy (small natural bin size).
- You want useful signal:noise levels (high fold and trace density).
- You want near offsets for good estimates of P-wave reflectivity.
- You want far offsets for good estimates of elastic properties (e.g. density).
- You may need far offsets to image certain types of geology.
- You want the acquisition crew to have enough receivers for the design.
- You want to be able to afford the acquisition and processing!
A while ago, I wrote a blog post on the subject: Fold for sale. It's about balancing fold (how many traces go into the stack, basically, a big driver of signal:noise) with cost.
My colleague Evan Bianco and I also did a series on modelling seismic acquisition. If you're into Python at all, there's some code to play with:
The greatest minimum offset in a bin is an important consideration, as you guessed. Also notice the spider plots, which try to visualize both maximum offset in a bin, and the range of azimuths going into that bin (which could be important for stress analysis, for example):
It's hard to go into a lot more detail, other than pointing at more things to read. There are a couple of really good books on this subject:
For a quick overview, I recommend reading this great paper by Norm Cooper: