In broad terms, the goal with ZS EI testing (fixed density) appears to be the exposure of shadow detail on or close to the straight line of the curve. According to Adams, important shadows should normally be placed on Zone III. For texture/detail to exist within an average Zone III value, we must have a range of values, say at a minimum Zone II to Zone IV. It then follows at a minimum you'd want Zone II on the straight line. In practice, with most current films, when you follow the typical Zone System EI test (meter, close down 4 stops etc, 0.1 above B+F etc etc), I have found this is more or less what you get. You can take this as far as you want. Barnbaum, for example, advocates using half the ISO speed and placing important shadows on Zone IV.
However the assumption the shadows need to be on the straight line is a Zone System principle. It is not based (as far as I know) on the first excellent print studies or fractional gradient speed method, where some compression in the shadows seems to be acceptable, as long as the shadow gradient is related in a specific way to an average gradient.
On the face of things, if you buy the first excellent print studies, it would seem ZS EIs give you a safety factor rather than improving print quality. Perhaps ZS EIs also give you some more flexibility when interpreting the negative. Whatever you want to call it. What do you lose by giving extra exposure? Under normal circumstances, not much (besides the loss of speed). Technically, overexposure results in increased granularity, irradation and halation (both contributing to lower acutance). But with current films are these effects even material? Probably not when we're talking a half to one stop.
It is important, however, that as you reduce development to accomodate scenes of a higher contrast range, the difference between ZS EIs and fractional gradient speeds increases. Since the fractional gradient speed is based on a shadow gradient in proportion to an average gradient, it is that relationship which must be maintained to produce excellent prints, not a fixed density. Clearly with a fixed density criterion (ZS), the more you reduce development, the more you have to reduce the EI. This doesn't appear to be the case with fractional gradient speeds, at least within a reasonable range. So, relative to the first excellent print criteria, ZS EIs in "minus development" scenarios result in increasing amounts of "overexposure". What are the costs under these circumstances (besides more speed)? Well, slightly more of the image characteristic effects stated above. What are the benefits? Unclear.