We should prrrrrrobably just delete Phlux.

I'd like to send these applications to the farm after T5055 (move them to some separate, explicitly unmaintained archive that we probably never touch again):

• Phlux - Almanac does similar/equivalent things better in most cases, and we never ended up putting this into production in the cluster, and don't really have much of a use case for it.
• Phragment - I think the feature set is conceptually fine, but this was a one-install project which has basically no interest or traction. If we built this in the future we'd probably be better off starting from scratch.
• Releeph - Feature set alignment (T9530) but Releeph itself is hard-coded Facebook workflow stuff and I think we're better off rebooting it completely.
• Phrequent (?) - I don't hate this one too terribly much but it doesn't seem like something that any customers use or are interested in. Maybe return it to the upstream some day.
• Ponder (?) - Ponder isn't that bad but it's not very good either. It does get at least some use.

If any of these have been obviously-broken for years (e.g., nothing in Phlux actually works or something) and no one noticed, I'd be inclined to just kill them outright.

Alternatively, this can likely be fixed by creating edge tables for Phlux and adding them to PhluxSchemaSpec (which it looks like does not exist yet).