Well, apart from the general principle that less is more, this is a specific kind of processing (dither with noise-shaping), and it isn't readily audible under normal recording and listening conditions.
But the seemingly magical ability to pour 18 or more "bits' worth of resolution" into 16 bits assumes that you will deliver the 16-bit signal in precisely the form that it emerges from the A/D; no level adjustments, rebalancing, filtering or EQ'ing can be done afterward and still preserve the full benefit of the noise-shaping.
With most live recordings that I make, some post-processing for level or balance has to be done--and at that point, of course, one has a 16-bit recording which must be redithered as such. Realistically, it would be better to record live at a greater bit depth--18 to 20 actual bits, for example--then use noise-shaped dithering to produce your final 16-bit tracks for CD (or whatever). That's where "UV 22," "SBM" etc. are at their most useful, I think.
--best regards