It's situational. When you need a search across multiple data sources, standing up a dedicated search engine makes a lot more sense. Then again...PG Foreign Data wrappers would make that scenario pretty simple without the need for an aggregate.

I can't speak to performance in that situation though.