There seems to be a divergence between a ttrpg folk tradition of spectacular crits and fumbles and what most actual game designers want to put into actual rules. This is most prominent in 5e D&D, but I'm pretty sure it's true of other games as well (ie: my Daggerheart group definitely makes critical successes a little extra-special good beyond just losing a stress and gaining a hope, as the rules say).
Frankly I think, from a design perspective, it's foolish to thumb your nose at people wanting to take joy in extreme results on their math rocks, it's clearly a strong hook for player engagement and the basis of many people's favorite rpg warstories. The simulationist math nerd in you saying "but 5% odds is too unrealistic" needs to be told that games are games and crits and fumbles are fun.
Now that doesn't mean crits and fumbles should have gamebreaking results. The optimal design is one that both satisfies the player who wants the tables exictment at a crit to be satsified but avoids casually handing out gamebreaking results. Personally for D&D I treat a 1 as about a -5 and a 20 as about a 25 on the die, and narrate accordingly. That doesn't seem to break 5e D&D, mileage in other games will vary.
In many games combat involves a lot of rolls, so a fumble on basic attacks should not generally have extra penalites as that is going to undermine a PC feeling competent in their adventurer (or whatever) vocation. And if some other game has some other sort of roll that is made comparably frequently I would advocate having a muted outcome for fumbles there as well, unless the premise of the game is that the PCs are bumbling idiots.
