@JayT Great questions! Going to tag @dc42 as backlash compensation is a newer feature (and I don't see it listed in the RRF change log!).
I have used a dial gauge and measured the backlash on my Y-axis, and measure about .15mm. With only a small amount of testing, it seems as if the compensation works well.
Intuitively I could see how a smaller value of "S" would improve more print segments, as 10*backlash=a fairly large value (relative to things like small diameter circles).
A) I would not use a 'calibration' cube to calibrate things like backlash or print size/scaling. I would only use a calibration cube as a quick check to ensure the printer is functioning at its lowest level (able to move, extrude, be at roughly the correct size, right height off build plate, etc). For true printer calibration, I highly recommend this video: https://www.youtube.com/watch?v=H7OsnMLDIMw I also highly recommend purchasing the model/spreadsheet, as it is really well done, and offers insight into extrusion, print shrinkage, and axis skew.
As part of that, I wonder if printing the model without backlash compensation, and then with backlash compensation, as merits. Something I could certainly test. As far as other methods of measuring backlash (you're using a dial gauge, which when used properly is probably the best method), here is an example on how an older professional style printer measured Y-axis backlash - a series of zig-zags, where the user inspects the print picking the most evenly spaced lines.
Something like this could work well:
https://www.thingiverse.com/thing:3060573
C) Axes having uncorrected backlash of differing amounts will definitely produce print defects. I can't imagine correcting them will induce more/different defects, however since this is a newer (and somewhat unpublished feature) reporting testing and validation to the forums will be of value to the RRF dev team.
D) I think the recommended value of "S" is going to vary greatly printer to printer. I could see on other printers (say a printer with some sort of rack and pinion) with a much greater backlash value might need an "S" value closer to 10 (As the amount of backlash needed to be compensated for could cause motor issues on shorter segments), but I think on my printer (Y-axis with some belted gearing down) the .15mm could likely be taken up easily with very small values of "S". It would be interesting to test this by running something like a 'ringing tower' or some other test print, changing the "S" parameter with height (decreasing "S" as the height increases), and noting if there are any visible differences.