Unable to replicate in testing using prod containers locally, so CloudFlare seems like the most likely option?
All of our downstream (Nginx + Gunicorn) have 3600s timeouts, so there's nowt left. Inspection of logs there reveal no obvious issues.
Their docs[1] say they time out at 100s, so that's spot on.
Workaround: Results are logged locally and in the database, so just inspect them.
Solution: We're already using RQ, so create a simple set of RQ jobs that run the loader and return the results.
Edit: Oh, wait, we'd already lodged a ticket for this: #52 🤦♂️
We're pinned to 2.2.1 until it's resolved.