For the big OP point of AD&D's reverse bell curve big impact ability score system I was quite unhappy with it by the end of 2e.
The disparity between a 17 strength fighter and an 18 percentile one is significant. Gatekeeping of classes behind stat requirements was annoying. The rich get richer phenomena was significant and broad, in bonuses, class qualifications, max level, max spell levels, roll under proficiencies, and bonus xp. Stats mattered.
I liked the human rolls chart in UA that allowed PCs to have great scores to make heroic action star characters and at least hit guaranteed minimums for the classes they wanted to play. I was conceptually glad when I saw Dragon came out with a similar system for demihumans even though I did not own that issue.
2e Players Option Skills and Powers added similar dice assigning methods and point buy options to craft how you wanted your character to be and had a neat option of splitting scores into subabilities that could diverge but had to be within 4 points of each other, so if you rolled a 16 strength, you could have an 18 percentile for combat but a 14 strength for bending bars and such. I thought this was great but it had contradictory statements on whether these were capped at 18s. I felt if they were uncapped it just made the rich get richer problem worse as going from a 12 strength to 14 gives no benefit, but going from 17 to 19 is huge.
Even with these options I was getting a bit dissatisfied with AD&D stats. I wanted characters generated randomly to be different and fun, not just hugely disparate in their effectiveness which I did not find as fun. NWPs using roll under stats also pointed out how much of a difference an 18 versus a 10 meant in an ability check (base 90% success versus fifty fifty) which really skewed things when different people tried the same general adventuring thing.
By the end of 2e I was starting to think that B/X Basic looked like a better option with the 13-15, 16-17, 18 +1, +2, +3 modifiers not being so huge a disparity if there was a difference, so 3d6 randomly rolled in order could be viable and not mechanically terrible in comparison to good rolls. Also how there was no gate keeping on classes and levels from stats.
Importing that to AD&D would mean top fighters would generally be a bit less powerful (no +4-6 damage from big percentile strength). It would also require deciding on whether to have minimums for classes like rangers or paladins or just let them be superior options for good characters.