That importing is expensive and prevents you from handling billion scale graphs.

It's possible to run cypher against duckdb (soon postgres as well via duckdb's postgres extension) without having to import anything. That's a game changer when everything is in the same process.