I write one Megabyte of data with appended newline characters. The virtual machine environment completes the writes in one second. The same machine without the emulation requires less than four seconds. Another Fedora machine required about four seconds, as well.
Is /dev/urandom producing significant variation in random bytes? The architecture is Apple M2 with a Parallels virtual environment.
It depends on how Parallels sets things up. No idea.
In general, once seeded, urandom is safe to use.
I am by no means an expert on the subject, but I found the following links informative:
A lot of work has gone into the RNG’s in linux kernel in the last couple of years to make it so that the answer is YES its random.
You can read a lot of the background on lwn.net that goes into the changes and the tradeoffs, especically in early boot.
A web search for “lwn rng” will provide interesting pages for your research.
I will bring your attention to this one Uniting the Linux random-number devices [LWN.net] that provides a lot of info on current kernel RNG implementation.