https://github.com/openzim/zim-requests/issues/1346 seems to indicate that we do not retry on failed HTTP calls. We should probably do this with a backoff strategy to not overwhelm upstream.