Fix avg & unique by reverting to old cartesian join

[pieterbeulque]

Just to test. Follow-up to #8769

To be correct, I think we should do something like this, but could run some numbers to assess the impact on performance etc?

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
polar	Ready	Preview, Comment	Feb 2, 2026 9:09am
polar-sandbox	Ready	Preview, Comment	Feb 2, 2026 9:09am

[frankie567]

@pieterbeulque Should we discard this?

[pieterbeulque]

… well. It's not calculating correctly in production right now, but merging this would be a performance regression. @Yopi - since you "broke" it / fixed the performance issue (half empty / half full kind of way to look at it, haha), what do you think?

Not pointing fingers 😊

Actually, not sure how it'd go together with the other optimizations made in main since this PR.

[Yopi]

I'm also not sure. Maybe we should merge this just to get it working correctly. Given the other changes we've made it might be fine out of a performance improvement perspective.

[pieterbeulque]

@Yopi - I rebased this against main so maybe we can merge it now? I think the issue is still in main.

[Yopi]

@Yopi - I rebased this against main so maybe we can merge it now? I think the issue is still in main.

Lets do it