I'm not sure I'd agree it's a fluke. Someone I worked with at a previous company did something very similar to this, and I recall reading an account on HN a couple years back about someone who did it using different criteria than OP, but still in a way that surprised me.
Be a machine learning engineer, have published one paper, have peer reviewed code, work for a company, be paid typical MLE comp, and you meet 4/8.
From my anecdota it seems like they won’t really question it.