@despopoulos_chriss, you make a good point about ROI.
We like DITAVAL for book-wide profiling, as product conditions are group-wide and stable. But this would not be true for ad-hoc usage. If I make DITAVAL files for "tableA" and "tableB" for one specific book, then later rewrite the content to remove those tables, now I have a dangling DITAVAL file to be noticed and removed.
DITA is mechanically capable of many, many things. Many features are readily accessible to writers in a DITA editor (Oxygen). Perhaps the more mechanically involved features require a CMS to orchestrate the moving pieces. I think what we're discussing here is, could DITA somehow support ad-hoc conditionalization via the former instead of the latter?
P.S. We also use a small fraction of DITA. We specialized it to match the elements in our previous format (structured FrameMaker). We use variables and profiling. We are just starting to explore con[key]ref. We don't use subject schemes, branch filtering, or anything like that.