Hm, I don't know about that. I think a lot of the objection to talent trees comes from the playerbase and perceptions on how the game 'ought' to be played, and it's certainly been proposed a few times before. Branching options can co-exist with optimization. You just can't have a clear means of comparing the two.

As an example, take a look at the tank 30% DR cooldowns. Vengeance is still the best of them all, just because it does everything the others do and has an added thorns effect. You could say that there's a 'mathematically correct' answer on which is best. Warcraft has min-maxing too, but its talent trees don't always have such obvious 'right answers'. Let's say that you have a gap closer that moves you forward a fixed distance (Roll). Do you want another charge of it? Do you want an upgraded version of the base action that gives you a temporary speed burst after (Chi Tornado)? Do you want a completely new action added in that lets you grant a movement speed boost to a single party member? You could argue for and against each of these in a fight-specific fashion, but you're not going to get a clear consensus on it.

There are other applications for this approach as well. The development team started to look into ways to simplify jobs starting from around Midas, and one of the first targets were maintenance effects. And rightfully so, because there were a lot of complaints coming in around upkeep buffs. But what if your choice is between maintaining an upkeep buff for slightly more damage versus having a passive that does the work for you for slightly less damage (but more than if you were inconsistent with the maintenance effect)? The latter is more consistent, and if you know that you struggle with that it may actually be the better choice. It lets you keep jobs accessible without having to compromise the more interesting design elements for players who like them.

I'm not arguing for or against this approach, but the reason why it is believed to 'not work' here is not because of the limitations associated with two minute buffs or even because it's actually been tested and failed. It doesn't work here because the community itself has very strict notions on what works and doesn't work, and that's a much bigger constraint than anything else the developers are faced with. If we were all a little more open-minded and a bit less resistant to change, there'd be a lot more room for creativity.