There has been a lot of discussion on this forum about the performance issues, e g
Performance Issues with Gramps, in particular Performance Issues with Gramps - #16 by UlrichDemlehner
Error using PostgreSQL as database backend with Gramps v5.2.2 - #14 by UlrichDemlehner and below
Since it was clear that those performance issues are not the consequence of a specific bug but the consequence of an old software design decision (using Python blobs), it made no sense to file a bug report.
An executive summary of my experiences (all statements refer to Win10, database on SSD drive, approx 90 k individuals with 350 k events):
(1) When migrating from v5.1.6 to v5.2.2 and from BSSDB to PostgreSQL and SQLite, I found that the performance of searches went south. Regex name searches in the person view that used to run for 15 - 20 sec (which by itself is clearly not very impressive) now needed 45 - 60 sec, searches in the event view (which have always been the slowest searches by far) went up from 45 - 60 sec to 100 - 150 sec. Even searches in the other views (e g citiation or place view) which used to be quite fast (around 1 - 3 sec) were now significantly slower.
(2) I haven’t noticed any significant differences between PostgreSQL and SQLite but I have not done any sophisticated speed tests comparing both database backends. Based on the discussion I refered to above, this was exactly what to expect so tests did not make much sense for me.
(3) So at the end of the day, I had to go back to v5.1.6 on BSSDB where I’m still. And to deal with those performance issues, I wrote some apps by myself which basically “import” a Gramps XML file into a PostgreSQL database where I do all my searches.