To answer the questions around source and weighting etc:

1: I'm using https://www.hnsearch.com/api, which the page itself seems to indicate is used by HN for certain views.

2: The url is http://api.thriftdb.com/api.hnsearch.com/items/_search?prett... The API won't let you fetch more than 1000 results in total (using limit and start in tandem).

3: I'm fairly certain the API is fairly new because I never came across any results older than 2010 while fiddling with it. As such the couple of suggestions of weighting by active users might not return the desired balanced result.

4: I'm fetching the results at one go (all 1000) from the API through a cron the runs once every 48 hours.