One of the succeses of the refined D&D class system is that many distinct mechanics and spells are good, so players can choose what is fun and not google "best build"
Though one of its biggest issues is that there's quite a lot of not as good stuff. Then you add on limitations to such things to promote "Different pathways" that just further pushes people towards good stuff.
In fact, it's a sentiment that's been brought up a number of times on these forums. Especially in relation to party size. Whereby things like spell selection pushes them towards the few good spells and there's not enough freedom to pick up or use more niche spells that might be fun.
I.e. Why would a Sorcerer who only gets 3 level 3 spell slots pick things like Daylight, Gaseous Form or Hypnotic Pattern when things like Fireball, Counterspell, Lightning Bolt and Haste exist?
This also ends up being true for classes and subclasses. Many people dunk on Rogue outside of Thief multiclass just for Fast Hands. Why would anyone play a Monk that isn't Open Hand? Why ever play Ranger when Paladin and Fighter (And Sword Bard) exist?
This isn't unique to BG3 either. Thoughout all editions of D&D there's been a constant struggle with class balance, especially in regards to classes like Rogue and Ranger and the absolute stack that is Fighter and their excess of bonuses (Where at times Fighters even pushed out casters because why play a caster and have to suffer through dogwater early game when even your peak is eclipsed by the Fighter swinging his Greatsword 20 times a turn infinite times per day?) and of course scenarios when magic is just utterly bonkers and a single high level caster just wins every encounter alone.
Balance disparities between classes can be a major issue. Both leading to "Obvious Best Builds" but also leading to the easy way out... The dreaded "Homogenization" band-aid. Where classes lose their uniqueness (Thus players their build choice) so they all have the same base kit and thus function around the same power.
The most clear example of this would be the Dragon Age series. Dragon Age: Origins had diverse builds and party compositions that really mattered... Though the result was that Mages are OP. By far. Not only have way more options than Rogue or Warrior but vastly more power to boot. Fastforward to Dragon Age: Wokeguard and classes don't matter, the only difference is what gimmick your ranged attack uses (With a side of resource generation disparity) and party composition doesn't matter because everyone does the same thing regadless of class. Everything is homogenized into pure nothingburgers.
A skill system that is too open can remove variability and diversity which makes campaigns feel repetitive and boring. So the class design needs to have some way of keeping characters distinct and allowing many different pathways to be effective.
At the end of the day it's all about providing many options, but having sufficient limitations.
The more options you have, the more diversity you can have.
While the limitations keeps things in check so you can't eat your cake and have it too.
This is the difference between "Good systems" and "Bad systems". A bad system won't have enough options or limitations so builds all end up very samey, either due to lack of choice (Not enough options) or a lack of restriction (Not enough limitation).
It's here that the aspects of "Class" system present itself. A "Class" system inherently has limitations through you only accessing what a class is pre-prepared to obtain. The cost is the lack of options, which can be mitigated by making a LOT of classes and/or having multiclass functionality, but still persists.
It's also where the ability for a "Classless" system to fail is shown. With no inherent limitations, these instead need to be specifically designed and implemented to make it succeed, else it becomes "Master-of-all-trades". But of course the benefit of proper design is unparalleled options.
