I am very much in favor of putting readable IDs into the paper version. Yes, for long term considerations, but also for short term considerations. I believe anything that is not there will not exist for the taxonomists and be considered only unintelligible techie stuff. And therefore certainly not worth thinking about in the context of their own work.
With respect to how to handle the presentation I think taking a look at how Genbank/EMBL sequence accessions are handled may be helpful. Not every occurrence of "Chromis" in the text needs the ID associated (it may have a hidden hyperlink though), but at one point it needs to be clarified. In molecular phylogenetics it is common to do this a single time in a table in material and methods. However, in the discussion an additional accession number may just be cited in brackets.
In short: allow any normal publishing practice, consider it as a special form of reference (like doi or ISBN) and observe the normal publishing practices of citing, especially avoid redundancies. Rich, I think you are too much thinking about general rules how to always handle it - but publishing practice for good reasons does almost the opposite (once and never again...).
Gregor