I think I'll go with the +1 damage per die every two levels, it might swing a bit too far the other way but it'll be interesting. So a 6th level Warrior with 5 Strength does 2d6+11 damage with a longsword, 2d6+14 damage with a halberd, and 3d6+14 damage with a two handed sword.
On that note, I suspect that the reason for stat gain slowing down as it increases is to encourage spending points on more than your accuracy and damage stat (Fighting and Strength for Warriors, Accuracy and Strength or Perception for Rogues, and Intelligence and Willpower for Mages). Does the system break if this cost scaling is removed?
It still breaks down. And adding a stat that does absolutely nothing other than be your attack modifier was a horrible decision for a game that recommends equal parts RP, exploration, and combat.
And yes, the system breaks down if you don't use ability advancements. Actually, it also breaks down if you do use them. All you really need is a +7-8 to succeed at most challenges.
Personally, we fixed it at my table by switching to d10s, although I wished I had gone to d12s (for easier conversion). We ditched ability advancements, the players need to diversify more (less autohit and auto fail), stunts are both stronger and more rare, and actions are more consequential (attacks rarely do a couple damage).
Because I didn't want to give tons more SP, the basic test roll is 3d10+modifier &1d6. The 1d6 determines only the degree of success and amount of SP generated.
If I had to start over, I would use d12s, and just double everything (longbow 1d12+6, Defense base 20, etc.). Then I could have had 1d12 be the action die and just halved it for SP/degree of success (1-2 is 1SP, and so on).
It's a little more involved than just lessening HP, but it really made combat more meaningful. A 2H sword with 7 STR would average 7.5 damage vs heavy plate; with my system that would be 3d10+9 (no ability advancement penalty) for 15.5 average. With Mighty Blow and Lethal Blow, it would be 18 vs 32.
I wondered if the ability advancements would lead to more min-maxing, but the opposite was the case in practice. The bell curve still places a soft cap, but with less auto hit and auto fail, it increases the need for diversification. Resisting spells for example, is actually possible with a larger spread, so more points have gone into defensive stats because it is actually possible - and not just pumping damage stats.
Honestly, I like stunts being pretty common. I'm going to be keeping to 3d6, I'm just going to bump difficulties and defences up slightly. I don't think the scaling is too bad, I'm unlikely to be running all the way to level 20 anyway but even if they focus entirely on attack they'll be missing any damage scaling, while if they focus on their damage stat they will start hitting less. If I increase defence at a rate of about +1 every four levels it should mathematically even out (although likely slightly below that to encourage spreading out your stats).Personally, we fixed it at my table by switching to d10s, although I wished I had gone to d12s (for easier conversion). We ditched ability advancements, the players need to diversify more (less autohit and auto fail), stunts are both stronger and more rare, and actions are more consequential (attacks rarely do a couple damage).
If we work on the assumption of abilities that are dumped need not be competitive at higher levels, we likely end up with a 4-6 range of target numbers for 'everyone has a chance'. +8 means you'll hit an 18 half the time. This means a legendary hero will beat formidable tasks in their primary areas of expertise a little over half the time, their secondary areas of expertise a little under half the time, and rarely anywhere else. This sounds about okay for a character meant to be closing hellgates rather than bring down the local bandit lord.
Of course, you can hyperfocus, and put all 10 primary ability increases into say Fighting, but at that point you've decided to dump everything else, and it's your fault if you can't reliably beat a hard check anywhere else.
Now you might not like how FA scales on abilities, I think it does okay, but I also don't think everyone needs to be able to contribute in every action. They shouldn't be left out in any situation, but there's no need for everyone to roll at that Formidable Intelligence (dire wallaby habits and diet) check.
Except you don't want people hitting slightly more than 50%, unless you are heavily altering AR and HP. It's not uncommon for an attack to do nothing.
And your example of TN 16-18 highlights the problem with scaling. That's great for the +9 character, but the +4 one? They have a 16-37% chance of success. So unless your group is all hyper focused, or fine with specialization, you will have parts of your sessions that are dedicated to certain players by necessity of auto-failing. And that opens its own can of worms.
Your mileage may vary, of course.
And you get enough focuses to pick several areas to be at +6 in by level 20, sure it won't be everywhere but it'll be enough to have an okay chance at those amazing feats. Why are we letting those who haven't focuses in those areas have a decent chance of passing DCs that are supposed to be hard? Isn't that the entire point of high DCs?