Jest test with seedrandom is unpredictable when source code changes, and test fails

I've created a small reproducible test case here: https://github.com/opyate/jest-seedrandom-testcase

When I use seedrandom, I get predictable randomness, as proven by the test.

However, when I add inconsequential code to the test, it fails.

Can anyone shed some light as to why this might happen?