Lemmy.world has been under repeated attack recently though, and the behaviors you’ve described match what I see when th service is down. You can see current status and the history of frequent incidents at https://lemmy-world.statuspage.io/.

To relate to your statement about what fails and how, I can say I’ve seen the failure-modes change as they adapt the setup, and it’s a more complex stack than other lemmy instances in order to deal with the attacks and large scale. It degrades in complex ways that are hard to fully reason about unless you’re pretty deeply familiar with how things are out together.

I suspect you’re seeing a combination of “lemmy world is broken sometimes”, “Cloudflare gives weird errors sometimes”, and “clients cache things or degrade to unauthenticated connections sometimes”. But in any case, seeing lemmy.world be flaky is not weird, it’s having a heckuva time.