My concern would be that classless and skill tree systems tend to produce much, much less variety, not more, as players converge on a few "optimal" builds. I think that's what you would see if you went to a blended subclass system.
That’s certainly possible. Though one might argue that this is a sign of imbalance…
Take the 2nd Figther level, for example. I suspect most people care for Action Surge and fewer people care for Tactical Mind (although it’s not bad either, mind you).
Let’s ignore prereqs for now to keep the thought experiment simple (or alternatively, assume that 100% of the 1st level Fighter’s abilities are prereqs to each of the 2nd level Fighter’s ability).
Let’s say each level grants you 10 points to spend, and we have a design constraint that Action Surge (once per short rest) + Tactical Mind = 10 points. How would you allocate the cost of the two abilities?
The "everything is a feat" approach values everything equally, which is equivalent to saying each are worth 5 points. In that context, I think it’s fair to assume many people would pick AS and save their other 5 points for another ability than TM.
But what if AS costs 9 points and TM costs just 1? Would we still get more characters picking AS than TM? Maybe, maybe not…
Another idea would be to include "package deals" that represent the benefits of training in abilities that have synergy with one another. Or put another way, the cost of flexibility. Under that approach, you get 10 points per level, Action Surge costs 9, Tactical Mind costs 3, but you can get AS + TS for 10 points.
Anyway, these are just examples, but I think balance is not insurmountable, though it certainly is difficult…
