Please change default blocksize to at least 1MB #4629 links to results table showing progressive slowing (somewhat), and then a wall where some SQL just doesn’t finish (for some reason) in a reasonable time.
I can tell you that during the time the SQL was hung, one vCPU was pegged.
My different slow query this morning showed (in Process Monitor) a huge amount of reads on database, which may or may not be normal, with specific reads (recognizable by offset and length) being repeated.
This reminded me of how SQL operations sometimes must scan (not search) the table as listed earlier, however it’s definitely not a definitive analysis. Maybe searching tables also generates same reads a lot.
I looked at it with Process Explorer, saw the SQLite thread (identified by its stack) using lots of CPU time, which wasn’t at all surprising because I was also looking at my profiling log showing a query in progress.
Note that I’m talking about a different slow query here, during an attempt to make my own slow test case.
Backups suddenly taking much longer to complete is me musing on SQLite tunings for larger databases, however I’m way out of my league with SQL in general and SQLite in particular, so thanks for @rsevero.