I've always thought that random number generators are one of the best examples of Hyrum's law ("all observable behaviours are part of your API"): once you release a random number generators that either uses a default seed or allows you to seed it, you can't ever change it, it's a huge breach of backwards compatibility. Imagine if you did a Minecraft style game that relied on the behaviour of some PRNG, and then you changed the implementation? The entire game will break. That's why GNU libc still uses a terrible LCG for rand() despite the fact that much better generators exist: they can't ever fix it, because srand() exists and people rely on it.
On the other hand, it's STUPENDOUSLY useful to have "default" random functionality in your core library, for the "just give me a random number" or "shuffle this array, I don't care how" users, who don't really care about the details. But if you do that: always seed it with some external entropy (current time or /dev/random or whatever), don't even allow users to seed it. That means you can improve it in the future, because users already can't ever rely on the sequence. If the users do want to rely on the sequence, they should have to specify the exact engine they want.
TL;DR: System.Random in C# should not ever have been seedable, big mistake.
I had a similar realization recently; I was writing a compiler so I implemented a "random" function as part of the runtime.
To avoid regression I have some simple code examples I compile and execute, and I compare their output to "known good" versions.
I reached a point where I wanted to write a "sort array" routine and my immediate thought was to generate an array of 50 random numbers, sort them, and print them. But of course that wouldn't give me predictable output for my test-driver.
In the end I decided I'd do that when run interactively, but for testing purposes I'd just sort the characters in a string "The quick brown fox .." and while it isn't super-convincing it's enough to let me see regressions in my sorting function and/or array indexing runtime code.
That's a bad example to use because it has very few repetitions (only the spaces I think?) and the key doesn't have different equivalent values so you can't test that you're order-preserving (or not).
But ideally sort is something you want to test with something like quickcheck/hypothesis, not gold tests (and I say that as probably the world's number 1 proponent of gold tests).
If you’ve played the game it makes sense to have the seed be settable and shareable. In Slay the Spire it can be exciting to have an outrageously unlikely starting state or early option, and in order for players to share this with each other the seed has to be user controlled. It’s a big part of what gives the game its community!
GP isn't saying you should never have seedable random generators, just that they should not be part of the standard library because then the API promise is no longer that you get random numbers but that you get a very specific sequence of numbers which fixes the implementation as part of the API contract.
They can certainly be part of the standard library, it's just that you have to make the programmer be specific: it's a bad idea `new Random(seed)`, because you can then never update the `Random` class, and you might get stuck with a terrible default forever. But having a `new Xoshiro256(seed)` is a fine thing to have in your standard lib, given that quality PRNGs are such a common need.
That split is exactly what .NET 6 did. The parameterless Random switched to xoshiro256*, but new Random(seed) stays pinned to the old Numerical Recipes subtractive generator so historical seeds still reproduce, and that legacy generator is the affine one whose first output is linear in abs(seed), which is the whole root cause of this bug.