I was recently asked by a colleague to explore more rigorously a related subject (effects of playing with near inset on progressive near zone performance), so I thought I'd share my conclusions with the group since we've been deliberating this issue.
Reduction in viewing zone. Every 1 mm of error in PD will result in a similar reduction in the viewing zone size to one side as well as an equal reduction in the effective binocular viewing zone size (the region of the viewing zones seen simultaneously with both eyes). The size of the near zone of a progressive lens varies significantly between designs, and will also depend on the distance from the fitting point, but the width of the 1.00 cyl boundary at 85% of the Add for a typical +2.00 Add progressive is around 10 mm. This point is close to the upper limit of the functional near zone, and is roughly equal to the minimum width of the near zone. Consequently, you can assume that 1 mm of variation in inset will reduce the viewing zone size by up to 10%. Errors in inset will have an even greater impact on the intermediate zone; every 1 mm of error will reduce the intermediate zone by up to 15% or more, depending upon the design.
Induced horizontal prism. The horizontal prism induced by varying the inset is generally equal to the product of the inset error (in cm) and the Add power (i.e., Prentice's rule applied to the seg). Excessive inset will create base in prism at near while insufficient inset will create base out prism. For a +2.00 Add, "flipping" the lens will result in roughly 2.0 prism diopters of base out prism (which may relieve some convergence excess or esophoria). This assumes that the optical center of each distance zone is at the correct distance PD. (As a side note, my colleague suggested that this induced horizontal prism may be why some wearers prefer abnormal insets.)
Unwanted astigmatism. This value, in particular, will vary widely between progresive lenses, particularly since it is not uncommon for progressive lenses to have a little unwanted astigmatism at the center of the near zone. For relatively small changes in inset (less than 2.5 mm), you can expect up to 0.25 D or more of unwanted cyl around the reading point. I would say 0.10 D per 1 mm would be a good approximation for a +2.00 Add, which is pretty consistent with the Minkwitz theorem. This error will have a much greater impact on the intermediate zone though, which is not only more narrow but also has more rapidly changing astigmatism.
There were some other, less significant effects (like slightly reduced add power), but these are the issues worth considering. Keep in mind that these approximations will depend on the individual lens design, Add power, and the wearer's tolerance of blur. I have assumed a "worst case scenario" for some of the rules in order to keep the math simple and the issues in perspective.