The documentation for 1.30_2 describes sqlite_use_immediate_transaction. Have you got a test case that demonstrates the described problem? According to my reading of the sqlite docs, the described situation cannot happen. You could only ever get a SQLITE_LOCKED error when the conflict arises within the calling process and never in the situation described in the module POD. SQLite locking would have to be fundamentally broken to exhibit the behaviour described in the POD. You should always get SQLITE_BUSY back in the situations described. Of course, you can always contrive within a single process to produce a SQLITE_LOCKED error. But that isn't what the POD says sqlite_use_immediate_transaction is for.