And/or evasion, and/or access to skills, and/or bonus feats, and/or whatever else. When I'm multiclassing I'm adding class features access to additional trees. Add a split for a tree isn't really much different than we currently do when we add a split for a specfiic PrE. The real difference is mostly in a few small low level enhancements.
I already made a bard AA from Elf on Lamma. 18 bard 2 ranger. I spent all of my AP in elf and arcane archer and didn't even touch tempest or sniper (which I had hoped to add some). I couldn't purchase everything I wanted in those 2 trees, let alone more. I also learned that a person needs to be level 25 for the AA capstone doing it through the racial unlocks, and noticed that mental toughness enhancements now adds to spell crit chance.
This is where I disagree. I clearly can't purchase a lot of higher tier enhancements in the 4 trees we have, let alone adding more. When we talk about moving abilities deeper into trees the reality is nothing is deep in a tree. Fusillade require 11 ap, a feat, and level 6 on live, and would lock out other PrE's for arti if they existed. Fusillade on Lamma currently requires 22 ap, no feat, and 4 artificer levels. The majority of those AP are spent on bonus to hit and damage with crossbows and extra action boosts and are mostly not really lost because I spent on enhancements I wanted instead of enhancements I didn't. The biggest difference is more class levels to play with, more feat room to work with, and a higher cost on AP to get to it.
In your 12/6/2 build example you have 2 PrE's, with whatever bonuses. We are likely going to see 12/6/2 builds and now possibly 12/5/3 builds based on the PrE line and tree structures. Ultimately, we still have things to splash for, even in 12/6/2 builds, and there is still incentive to do so. I can make an 11 ranger, 5 cleric, 4 artificer under the tree system and spend most of my AP in cleric for the healing boosts, use ranger for my archery feats and manyshot, and use artificer for fusillade. 22AP+ in artificer and 42AP+ in healing domain and we only have 16AP- to spend in elf and whatever our 3rd tree is regardless. We already have 4 trees from which to select for that 3rd tree and we're looking at 3rd tier enhancements at best for one final tree.
Replacing any 3 classes and any 3 trees produces the same results when it comes to AP costs. That's why 3 trees isn't the limitation; there's no AP to spend in more trees once we've gone anywhere in what we already have and we can't even fill half of what we have. At best we can ignore top tier class enhancements and get up to 4th tier in 3 our of 4 of the trees we have.
The only thing I could see coming from the not having had the 3 tree limit is the ability to give up the strongest enhancements so that we can have more of the weakest enhancements because we need to invest AP heavily to get anywhere near the best enhancements, and we can only get to them in 1 tree as is. Choosing to spend AP on weak enhancements over stronger enhancements just seems counter productive instead of providing any real benefit.
I find the unlock costs to be very restrictive.
I think you are right in that we don't actually know what will be in those other trees and I agree that not all existing builds are going to come out of this as effective as they have been in the past.
I wrote this while busy, so if anything is seems a bit disjointed or train of thoughts trailed off I'm blaming others.