Remember this bugger? An annoying mon, to be sure. Serene Grace and iron head/zen headbutt, on a mon with 100/100/100 bulk and 492 effective speed, was a pain in the ass, if you weren't prepared. And even then, if you got unlucky, most of Jirachi's so-called "answers" could and would get flinched to death, even if they resisted stabs or had high defense. The only reason why it took so long to ban in the first place was that it's main counter just happened to be the best pokemon in the meta (Gyarados-Mega), giving the illusion that it wasn't a problem. However, not even taking into account the fact that this required teambuilding with Gyarados-Mega in mind if you wanted the best odds at winning, the fact that Jirachi mainly relied on getting the odds in it's favor
It doesn't matter if a strategy is massively effective, if the core mechanic of winning a game is designed around random chance, it has no place in a competitive game. There's no strategy in rolling a die.
And, frankly, Jirachi wasn't even as bad as some of the most notorious sleep users. Taking Jumpluff as example, a bad Jumpluff player will win any matchup it outspeeds, baring disadvantageous type matchups and specific counterplay options, about
55% of the time (assuming wide lens, assuming they don't sub after successfully landing their sleep powder, which is what should be done in most scenarios) (.83 x .99 x .66), and it just gets more unlikely for the person fighting sleep to win if the sleeper subs after sleep (Honestly I don't know how to calculate the exact odds of winning vs sleep after a sub, but you're basically gambling on you to get the 33% first turn wake continuously and him missing the 83% sleep powder once. Not very good odds). And if you look at basically every yawn user, like Snorlax, Relicanth, and Camerupt, you need a first turn wake to win, as you can attack turn one, do something turn 2 while they protect, then hope and pray, meaning they win
66% of the time. Hypnohex user Gengar-Mega, thanks to his gross base speed and special attack, just needs to hit hypnosis in the first place to win most of the time, or not get a first turn wake, bringing his chances of winning to be between
60% and roughly 40% (for when it's a 2HKO)
. All of these odds are closer to coinflips than consistent strategy.
Jirachi, on the other hand, needed a 60% chance to happen multiple times in a row, as for the most part it had trouble 2hkoing anything in the meta, and so on. Getting one flinch in a row was going to happen
60% of the time, getting a second,
36%, getting a third,
22%, and so on. Worse or comparable odds to the sleep users mentioned above. And as I said, it wasn't 2HKOing anything it wouldn't have already if it didn't have it's broken ability, so usually it wasn't going to win the highest 60% of the time.
Despite the relatively low odds of Jirachi cheese, council voted for a Jirachi ban after the community barely missed out with a 59% majority/non-super majority. The official reasoning given for the ban literally states that it was banned
"Because an entirely hax-reliant niche has no place in a competitive metagame,", 'hax' just being another word for RNG. The same logic applies with sleep.