Deterministic pseudorandomness is really useful for seeding dummy data, data that you’d like to be fairly random (though it doesn’t need to be high-quality randomness, so LCG is fine) but consistent across runs/page loads.