The point isn't to brute force in attempting to crack a single address, it's to use an algorithm that generates random private keys and the possibility is still there that on the very first try the key can be one that is tied to a large bitcoin address. I've done it and am doing it and have generated some addresses that match with already created addresses that have been used and a couple that have had small balances in them so it's not like it never happens. The only thing that makes it not so noteworthy is that it hasn't been tied to an account with 50+ btc or something, but the odds that I and others even reached one with an already used balance proves that private key collision is a common occurrence even given the trillions to 1 odds.
I find this really hard to believe, especially since you imply it's happened more than once. Are you generating truly random numbers, between 0 and 2^256? There's (almost) no way you could coincidentally generate a random private key which matches a funded address. Can you provide proof by showing the addresses and a signed message?