[In my experience] Sometimes that “flat” table design is the only feasible way to query and use the Bubble database. You can’t (without very messy subqueries) do the typical SQL approach of x.y.z property. So your traditional normalised relational database design simply doesn’t work.
Around performance under the hood - everything I know I learnt from this The Ultimate Guide to Bubble Performance - the new edition is out (now 210 pages!)
FWIW what I’ve found is however Bubble does it’s indexing - it seems to work. (probably because you cant access 2nd order properties like x.y.z I guess) so if you table has 10 rows or 20k (thats the biggest table I’ve dealt with) performance seems unaffected.
I’d be interested to hear anyone else’s experience. Beyond 20K + rows I suspect I will be looking to (optionally) use an external database like Firebase - search for @jared.gibb if you want to know about using external databases.