SQlite "merely" assumes that the problems that come with distributed systems are handled at the application layer. You'll have to solve those problems for yourself, sure, but in practice I have rarely (I think never actually) had dataloss through a fault of sqlite.

Also, did you know you can use in-memory instances (and share them across threads!) with the right incantation? And that you can backup your on-disk instance to an in-memory one, do your expensive transactions without hitting the disk then backup the modified instance right back to disk, even in-place if you want!

Sqlite is amazing when you don't expect the DB to do replication or failover on its own.