Had a group that would play DnD 3.5, where you need to roll to confirm crits (20 auto hits, roll again against AC to crit). We ended up rolling to confirm fumbles as well because catastrophic failure doesn't just happen 5% of the time. Imagine 5% of your army accidentally chopping their foot off or beheading their nearest kinsman every few seconds.
That's a 0.25% chance. Seems too low. I'd just repeat the test and if it results in a failure, it's a critical. That way the difficulty of the test would factor in.
I've always had beef with crit failures where it not only isn't balanced the way crit successes are. In other words, they don't really have a place outside of combat, and even there they should be not only rare, but realistic.
The worst a crit fail should be is giving the enemy an attack of opportunity. That's it. Maybe you could stretch it into something like dropping the weapon, or hitting an unintended target that was realistically in line with the weapon's movement, which still doesn't mean it would do full damage.
I dunno how many of the DMs that use it have actually been in a real, life or death fight, but it can't be a lot. I would say it's near zero just because of the stories people tell about crit fails in combat.
Strangely, where it would apply more is in skill checks that wouldn't normally be bothered with. But that would still be situational, like a carpenter hitting their thumb with the hammer. It wouldn't typically end up with some kind of devastating effects 5% of the time, either. Even when it did end up catastrophic, the fail on that kind of skill still wouldn't mean total failure often enough to merit it being a meme level thing. Even a drunk carpenter that hammers his thumb and needs to see a doctor can come back and finish the rest of the job. It might not sell, but the table would still be a table, if you see what I mean.
A nat 1 as a meme is awesome, I love the humor of it and play with it myself. But in game it not only isn't RAW, it isn't balanced out the way it sometimes gets implemented.
It's home brew to begin with. Crit fails were never part of official rules. It wasn't until 5e (iirc, might have been a line in 3.x) that it was officially supported as a known table rule that was in line with official rules. That makes iconic a misnomer for sure.
Also, as this thread should tell you, beloved is not accurate across the board. I would even call it controversial to some degree. There are players and DMs that are emphatic about not playing with it in place.
Fun? It can be. But it is also very commonly abused and done FAR outside of the usual range of possibilities for a crit success.
Truth is, on a d20 system, it just happens way too often unless it's dealt with the same way as crits of any kind are. Same with a nat 20 for that matter, when it isn't an actual combat crit roll. You have to roll to confirm crits because they aren't supposed to be that common. Why should that not apply to other forms of the idea?