As others have said, Working As Intended. Probably. Definitely Working As Designed. Whether the design team *intended *to be as sadistic as this process is, is a completely different issue.

It's like giving little kids knives to play with. Sharp ones. They get the general idea of it, but neither the kids nor the idiots who gave the kids the knives in the first place are thinking about issues of time or scale. What a 20% chance to succeed in an RE means is an 80% chance you will fail precisely the next time you try. It never includes two tries later, or three, or ninety-three. 93 fails on a 10% item is what I got once -- I've had worse, but this was the best for actual record keeping.

To get a string of 93 failures at a 10% chance in a row is so improbable, it calculates out to a 56 in 1 Million chance. That's a binomial probability for a string of related events, not the simple probability of your very next event. This game does not even have 1 million players to try to simulate that, that is how bad things can get.

But the devs just keep on playing with knives, not knowing the first thing about them.

56/1m = 1 in 18,000, which isn't that hard to believe given the game has held 1m subs for multiple months.