how does storage sync work?

[mrdrprofuroboros]

Hi there!
Can you please explain how storage sync work exactly? The problem I'm facing is:

• I've installed a self-hosted AFFiNE and connected to it via a web app
• I've stress tested it by creating a couple of documents containing the full "War and Peace" :)
• AFFiNE started lagging significantly, barely usable
• I stoped containers, purged my storage and postgres and restarted them back (expecting to see an empty workspace), but it got populated once I loaded it from the same browser
• I realized that there's 238MBs of IndexedDB app storage (a full copy, I guess) in the browser
• I tried the same stress test with your cloud + electron app (hoping it would have a somewhat better performance with a more robust local storage), but no luck, it slows down to a barely usable

Which is rather bizarre to me. Is that an actual architectural decision to have a full content copy? this feels kinda off for somewhat elaborate knowledge bases if you already have a centralized source of truth on the server. Is it possible to sync locally only the hot cache / actively modified part? or at least understand where the performance issues come from? e.g. obsidian doesn't care much if I work with a hundred of "War and Peace" copies (though obviously all of them are local)