unit tests never seed the RNG

[mjg123]

meaning calls like h.NewFuncName() always return the same values. This can lead to data leaking between tests.

[zootalures]

This shouldn't be a problem unless:

• tear down is not run completely
• you are using the same Fn instance for multiple concurrent runs of the tests

it should be safe to use a non-deterministic RNG (e.g. crypto/rand)

[mjg123]

It comes up during the init-image tests as there is an init-image generated during the tests which gets called randName()-init, so it's the same on every test run. It's possible that one test run could fail to build that image and end up using an image from a previous run, but I'm not considering this a high priority really.