>For example, to build shared threads like OpenAI’s, you might end up creating a new collection, a caching layer, or an index just to connect messages that logically belong together.
Which messages belongs together depends on the context and when you discover that the fancy thread store you built your app around can't return a list of MY messages without a full sequential scan of every single thread you'll wish you could solve it with an index.
Parkytowers is site about repurposing thin clients of various kinds, it's a goldmine for finding out power consumption, Linux compatibility, possible hardware mods, etc: https://www.parkytowers.me.uk/thin/hware/hardware.shtml
Around 20 years ago when I was trying to decide on what database to use my requirement were that it should store data reliably. I learned that MySQL in contrast to PostgreSQL:
1. wasn't ACID
2. didn't have foreign key constraints
3. could loose/corrupt committed data (no WAL)
Despite me not knowing much about databases it seemed like an obvious choice.
Which messages belongs together depends on the context and when you discover that the fancy thread store you built your app around can't return a list of MY messages without a full sequential scan of every single thread you'll wish you could solve it with an index.