You can of course give Bob membership in the Knights who say Ni and that nice magic crystall ball to the wizard, mixing and matching boons and items. That can be especially useful, if some players enjoy items more and some more are more into "internal" stuff. I really enjoyed the flexibility in 4e. But yes, it's extra.
We can do what we want, tho personally I wouldn't still use boons vs magic items, because then I might end up having two problems if I have to try to force some balance of boons vs items, especially when it's the same players who enjoy both.
Note that in the PHB, Shadowdancers are explicitly mentioned as Monks with the Way of Shadows sub-class.
It's ok, but that shows what we have lost in the process: in 3e the Shadowdancers could be a diverse groups of characters with different base classes, in 5e they all have to be Monks. The new mechanics have forced the narrative to change. Being tied to one class only is exactly the limit of using subclasses (although we can think about designing some subclasses so that they work fine on multiple base classes).
I've always loved the concept for PrCs for fluff reasons, but their implementation was too fiddly for my taste.
But that's pretty much what I loved about prestige classes in 3e: the character concepts that they often offered (excluding typically those PrCls designed to "fix" weak multiclassing combos).
The other thing I loved, is that they also introduced several cool functional/mechanical features, but sadly this was only sometimes. Many other times, they offered only a re-hash of existing features, or (even worse) generic "bonus feats".
I agree that implementation sucked, more often than not. And that was occasionally because of poor design, but IMHO the true reason for implementation suckage was that the 3e framework around almost required them to suck... With that I mean the following two issues:
- 3e prestige classes needed prerequisites that forced a minimum entry level, and some general overall "cost"; this largely dominated over narrative requirements, and turned them into a mini-game of "character build"; the original designers (i.e. Monte Cook) intended prestige classes open to as many base classes as possible, hence his original suggestions to avoid explicitly requiring a minimum level in a class, and use implicit requirements instead (such as BAB or skill ranks); however, most of the prestige classes afterwards were still designed specifically with one (or two) base classes in mind, so we got these pathetic wanna-be-implicit-but-really-explicit requirements
- because all characters needed to keep progressing in BAB, skill points, hit points, saving throws, and spellcasting, each prestige class needed to specify all these (plus proficiencies); this made their design a bit more burdensome, but most importantly it made a prestige class work much better with some base class and much worse with others, thus diminishing the original intended openness even further
I say that in 5e we are mostly
free from these 2 problems. There is still a need to define hit points, but it isn't a major problem. And we can totally drop mechanical prerequisites... those made sense in 3e also because the edition valued "system mastery", but as a secondary effect they also entitled players ("What do you mean I can't be a Knight of XYZ? It took me 6 levels to gather all requirements!"). If we moved entirely towards narrative requirements, maybe the players attitude towards prestige classes will also improve.
