Tim McCormack

New finding: Simply logging in to the UI seems to be sufficient to get notifications going again (for that user).

Oof, right. We actually got bit by this on edx.org, and what we ended up doing was moving the preview site (`preview.edx.org`) to be under the LMS (`preview.courses.edx.org`). I guess...

Hmm... @kdmccormick @ormsbee how hard would it actually be to make this change? It looks like everything reads `PREVIEW_LMS_BASE` via [in_preview_mode](https://github.com/openedx/edx-platform/blob/db32ff2cdf678fa8edd12c9da76a76eef0478614/lms/djangoapps/courseware/access_utils.py#L102) on the LMS and generates URLs via [get_lms_link_for_item](https://github.com/openedx/edx-platform/blob/db32ff2cdf678fa8edd12c9da76a76eef0478614/cms/djangoapps/contentstore/utils.py#L130) on...

I don't know that we usually take non-critical backports to older releases, but the diff is basically identical and so from that perspective I'd be happy to give an approval...

I'm going to see if I can [work around the caching issue on master](https://github.com/openedx/edx-platform/pull/31080), and if that works, maybe you can backport to Maple in a separate PR to unblock...

Oh sorry, I don't mean another PR for this work -- you can keep this one. I just meant that the change to using pip-sync would also needed to be...

I'm not fond of the idea of creating more work for people who are already doing the right thing. Maybe if it was _really easy_ to delete the marker? Or...

It looks like test errors on more recent PRs have resolved, so if this branch still contains pending work you may want to rebase.

I share Braden's concerns -- this seems like a really unusual use-case, and I'd want to have a clearer picture of why this can't be solved with table lookups or...

This can be closed out now, right? Looks like commitlint is now in effect on master.