The noise coming from an open microphone has enough entropy for what we’d usually assume to be random.

It is possible to hit a bottleneck getting random numbers, but there’s /dev/urandom which shouldn’t block it’ll just keep outputting without enough entropy to guarantee pseudorandomness.

Usually you’ll know if you’re hitting one of those bottlenecks, probably it’ll only come up in production or hopefully in load testing.

Yeah as long as the setup for the device symlinked its device path to /dev/random or you did that yourself. Stuff I used had a Java SDK

Yeah, not USB but in principle it’s the same whether it’s PCI or whatever. There’s no way to prove an RNG is truly random really just have to rely on validating subsets of its output look random. NIST has published recommendations for sampling RNGs.

Here’s the title:

A Statistical Test Suite for Random and Pseudorandom Number Generators for Cryptographic Applications

I don’t think there’s a real need unless it’s regulated to need some dedicated HW RNG.