And now a quick note based on my 40+ years of programming computers.

Cyrypto-quality random number generation, the kind that, say, Google would use to create root key pairs for their crypto-based certification service, is very very hard. Any substantial variation from perfection makes the keys easier to crack. Amateurs need not apply.

But game-quality psuedo-RNG is dead simple. Knuth published a simple and adequate algorithm back in the 1960's, if you insist on rolling your own, but why bother? Every major programming language I know provides an adequate and dirt-simple to use pRNG function either built-in or as part of a standard library.

Could SE have screwed it up? Of course, there's nothing you can't screw up in software. But the odds that they've accidentally screwed it up in such a way that a reported 89% chance isn't an 89% chance is very low, and I don't see a reason why they would do it intentionally.