For transparency, and as already hinted at in Who is using `paragraph` for queries? I intend to spend the next few weeks primarily on optimizations, specifically on indexing.
In current main indexing is split into two phases, but this is proving problematic. I’ll likely revert this, but rebuild a lot of the indexing in a different way with proper test and ability to benchmark it better. I think this is worth pursuing. I hope to cut down the “client bootstrap” time significantly this way, but we’ll see.