Ok. Thanks for the reference. I was not remembering the specifics on the 5.1 vs. 5.2 performance hit. Which seems to be about the BSDDB backend having much more caching allocated than the SQLite backend.
(Although the bogged down discussion on Geneanet about trying to troubleshoot the massive use of memory is troublesome. Expanding caching would probably just aggravate their problem.)
And @Nick-Hall suggested indexing could make an additional improvement.
Is that an accurate summary?