Ask HN: How do Hacker News articles become “Dead”?

I'm using the public BigQuery Hacker News data to track popular keywords and I'm thinking that filtering out "dead" articles would be helpful, but I want to understand how articles become "dead" first. Does anyone know what the process is?