Imo it's a result of the fact that 5e's Bounded Accuracy isn't actually bounded in both directions (for skill checks). It's fairly easy to get enough bonuses to automatically beat moderately easy rolls through Expertise, Guidance, and/or Bardic Inspiration. Conversely, it's more rare to be in a situation where a natural 20 wouldn't already be a success (partially because of the aforementioned buffing). Thus, in a system where Nat 1s/20s result in auto fails/successes on skill checks, characters are punished more often than they are rewarded. Particularly because a Nat 20 on a skill check (in BG3, afaik?) doesn't get you a *better* result than a normal success.
I run a lot of PnP games. One of the games I run uses the d20 Modern rules (which are based on D&D 3e). One on the PCs has a total intimidation bonus of +14. If he walks into a bar and wants to intimidate an NPC, unless the NPC is Count Dracula, the PC is going to intimidate everyone he chooses, no rolls required. What this leads to is a lot of role playing. Again, that's not a consideration in a computer game. The whole notion of "no matter how good your character is, there is always a 5% chance to fail" irks me.
Some GMs react to this by bumping up DCs, which is cheap because it negates the PC's improved abilities
