Whenever you encounter a bug, it's evidence that CI didn't catch the problem, and very likely it's not being tested at all.
That's a good chance to increase test coverage.

and what should the test be?

1. print "Hello World"
2. check that the username is the configured username

Which test would be failing currently?

I believe "Windows" would fail the username check, and always appear to be "System". There's an open PR now to address this. Is it difficult to include ephemeral windows runners in current test suites? It should not greatly increase costs, since they are ephemeral.