Right now, we run a lot of tests that ping external resources (GitHub, figshare, Zenodo, etc) on every CI run. figshare in particular seems to be blocking our requests on a regular basis, probably as part of a fight against AI crawlers. It would be best if we ran offline tests only on most of the CI jobs and allocate only 1 or 2 to run the full online test suite. We already have a pytest marker for that so it would be mostly a case of messing with the CI jobs.