I will leave the posed questions to others better able to answer them. I write to clear up a misconception you have made apparent. With respect to how SQLite arranges for DB integrity in the face of possible overlapping accesses, there is no practical difference between connections used one per thread among separate processes and connections used one per thread within a single process. So your stated reason for eschewing certain SQLite strategies for protection against concurrent access is fallacious. As a syllogism, it rests upon a false premise.