After discussion with Emmanuel, we thought that maybe the best way to fix the issue would be to generate either js files or json files (something we can easily consume in javascript) containing the entries when we build the in.relation.to website.

We would have to generate one file per project.

If it's not feasible, I think getting the (enter a number that makes sense here) last entries would be sufficient as we wouldn't need a lot of information in the stream (the date, the title, the url, a snippet and the tags) and we would filter them on the .org site. Having very old posts for shards for instance doesn't look like a feature so I don't think we would need to go back too far in time.